--- title: "List Conversations" method: GET path: "/v1/conversations" tags: ["V1 - Conversations"] --- # List Conversations `GET /v1/conversations` ## Request - Base URL: `https://api.app.noxus.ai` (the document declares no server; this is the origin it was published from) - URL: `https://api.app.noxus.ai/v1/conversations` - Auth: API key in header `X-API-Key` ## Query parameters - `order_by` string - `sort` string - `assistant_id` string, uuid, nullable - `page` integer: Page number - `size` integer: Page size ## Response `200` Successful Response - PageSdkConversation - `items` SdkConversation[], required - `id` string, uuid, required - `user_id` string, uuid, nullable, required - `api_key_id` string, uuid, nullable, required - `name` string, nullable, required - `created_at` string, date-time, required - `last_updated_at` string, date-time, required - `etag` string, nullable, required - `settings` CoWorkerSettingsOutput, required - `pii_banner` string: Information about PII anonymization - `pii_info` string: Information about PII anonymization - `pii_lock` boolean: Information about the guardrails - `pii_anonimization` object[]: Preferred anonymization method for PII values. - `divider1` string - `guardrail_info` string: Information about the guardrails - `guardrail_lock` boolean: Locks the guardrails settings. - `guardrail_latency_banner` string - `violence_and_hate` GuardrailValue - `value` string - `context` string - `sexual_content` GuardrailValue - `value` string - `context` string - `guns_and_weapons` GuardrailValue - `value` string - `context` string - `suicide` GuardrailValue - `value` string - `context` string - `crime` GuardrailValue - `value` string - `context` string - `toxicity` GuardrailValue - `value` string - `context` string - `bias` GuardrailValue - `value` string - `context` string - `profanity` GuardrailValue - `value` string - `context` string - `guardrails_scan_output` boolean: Scan the model's response with the same semantic guardrails applied to input. Catches sensitive content sourced from Knowledge Base documents or tools that the input scan cannot see. - `guardrails_output_mode` 'post_stream' | 'pre_stream': How output guardrails are enforced. 'post_stream' streams the response in real-time and warns/clears after the fact. 'pre_stream' buffers the full response, validates it against guardrails, and only emits it if it passes. - `divider2` string - `guardrail_action` union: Choose the action to take when a guardrail is triggered. - object - string - `guardrail_action_replace` string - `guardrail_action_human` string - `block_after_violations` integer: Block the chat after this many turns with guardrail violations. Set to 0 to never block (per-turn warnings still surface). - `block_message` string: Message shown to the user when the chat is locked. - `divider3` string - `prompt_defense` string - `prompt_lock` boolean: Locks the prompt defense settings - `jailbreak` boolean: Prevents attempts to manipulate or "jailbreak" the model into ignoring safety rules or system instructions (e.g., prompts like "ignore previous instructions"). - `prompt_injection` boolean: Detects and blocks strategies where users attempt to get around restrictions indirectly (e.g., asking the model to 'pretend' or speak as a fictional character to evade policies). - `context_leak` boolean: Prevents the model from unintentionally revealing internal context, instructions, or system messages that are not meant to be shared with the user. - `auto_model_pick` boolean: AI selects the optimal model for performance, cost savings or length - `auto_model_preference` string - `model` string[]: Model configuration - `parameter_preset` 'default' | 'creative' | 'precise' | 'custom': Controls randomness, diversity, and repetition of the model's output. - `intro` string - `basic_params_section` string - `max_tokens` integer: The maximum number of tokens per request - `temperature` number: The temperature of the model - `top_p` number: The top P of the model - `top_k` integer: The top K of the model - `timeout` number: The base timeout of the model - `divider_7` string - `penalty_section` string - `presence_penalty` number: The presence penalty of the model - `frequency_penalty` number: The frequency penalty of the model - `stop_sequences` string[]: The model will stop generating text when it encounters any of these stop sequences - `divider_8` string - `reasoning_section` string - `enable_reasoning` boolean: Whether to enable reasoning - `reasoning_effort` string: The reasoning effort of the model - `divider` string - `reliability_section` string - `max_retries_per_model` integer: The maximum number of retries per model - `retry_on_status_codes` integer[]: The status codes to retry on - `datacenter_fallback` boolean: Whether to fallback to a different datacenter of the same model if the current one is down - `exclude_user_identity_from_prompt` boolean: Omit the user's name from the system prompt. Keeps the cached instruction prefix identical across users so it can be reused, and avoids sending user identity to the model. - `enable_prompt_caching` boolean: Cache the stable instruction prefix at the model provider to cut input cost and latency on repeated turns. Disable to send every request uncached. - `protect_instructions_in_preview` boolean: Deployed conversations always keep the agent's system instructions, persona, tools, and configuration confidential. Turn this on to also protect them in the preview/dev page; otherwise the safeguard is dropped there so builders can inspect the agent they are configuring. - `usage_limits_section` string - `usage_limits_maximum_request` integer: The maximum number of requests - `usage_limits_maximum_tokens` integer: The maximum number of tokens - `usage_limits_max_messages` integer: The maximum number of messages - `usage_limits_agent_timeout` integer: The maximum number of seconds the agent can run for - `divider_11` string - `compaction_enabled` boolean: Automatically summarize old messages when context limit is approached. - `compaction_strategy` string: How to compact chat history when it grows too long. - `compaction_threshold` number: How full the context window may get before compaction runs (0.01 to 0.95). Lower compacts sooner and more often. - `compaction_tool_enabled` boolean: Expose a tool that lets the agent request compaction manually. - `tool_calling_section` string - `parallel_tool_calls` boolean: Whether to allow the agent to call multiple tools at the same time. - `tool_timeout` number: Maximum seconds a single tool call can run before timing out. - `visuals_section` string - `hide_reasoning` boolean: Whether to hide the reasoning messages - `group_into_chain_of_thought` boolean: Group reasoning and tool calls into a collapsible chain of thought - `chain_of_thought_final_answer_only` boolean: The agent only responds at the end, keeping all intermediate messages in its chain of thought. - `hide_workflow_progress` boolean: Whether to hide the flow progress - `hide_error_banners` boolean: When a turn fails on the web widget or an embed, hide the error banner from end users; the agent's own reply still explains what it could not do, and a turn that fails before it replies still offers a retry. The platform's own chat always shows the error while you build. Slack, Teams and the other channels never rendered these banners. - `tool_display_preset` string: Controls how tool executions are shown during conversations - `minimal_caret_only` boolean: Show a loading message while the agent is working. - `message_footer_enabled` boolean: Show a fixed footer under the agent's final reply in every channel. - `message_footer` NoxusTipTapDocumentOutput[], nullable: Shown under the agent's final reply in every channel. Never sent to the model, so it cannot be reworded or dropped. - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - `type` string, required - `attrs` object, nullable - `content` NodeOutput[], nullable - `marks` Mark[], nullable - … - `text` string, nullable - `minimal_tool_message` string - `citation_style` string: How inline citation references are displayed - `deduplicate_sources_by_url` boolean: Group sources by unique URL in the sources panel - `omit_citation_quotes` boolean: Hide quoted text in citation bubbles, show only the URL - `show_source_images` boolean: Show document page images presented by the agent expanded by default instead of collapsed behind the source images toggle - `chat_input_placeholder` string - `divider_visuals` string - `transport_show_tool_calls` boolean - `transport_show_sources` boolean - `transport_show_thinking` boolean - `transport_allow_noop` boolean - `transport_thinking_text` string - `conversation_title_section` string - `agent_title_mode` 'generate' | 'first_message' | 'last_message' | 'custom': Configure how the agent chat title is displayed in the chat history - `agent_title_prompt` union: Custom instructions for the generated conversation title. - object - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - string - `divider_conversation_title` string - `feedback_section` string - `feedback_placeholder` string: Placeholder shown in the inline feedback box. - `feedback_submit_label` string - `feedback_cancel_label` string - `capability_versions` CapabilityPins: A version per capability. Absent means "whatever that capability says production runs", which for an opt-in one is off. - `is_coworker` boolean: Whether the agent is a co-worker - `preset_selector` string, nullable: Choose a preset to get started quickly. - `languages` AgentLanguageOutput[]: Welcome message per language. The deployed widget's language picks one, otherwise the visitor's browser locale; the default language is used when none matches. - `code` string, required - `introductory_message` NoxusTipTapDocumentOutput[], nullable - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `default_language` string, nullable: ISO 639-1 code of the language used when the visitor's locale matches none. - `persona` NoxusTipTapDocumentOutput[], nullable: Describe who the co-worker is: their background, role, or point of view. - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - `type` string, required - `attrs` object, nullable - `content` NodeOutput[], nullable - `marks` Mark[], nullable - … - `text` string, nullable - `tone` NoxusTipTapDocumentOutput[], nullable: Define how the co-worker should communicate: its voice, attitude, and style. - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - `type` string, required - `attrs` object, nullable - `content` NodeOutput[], nullable - `marks` Mark[], nullable - … - `text` string, nullable - `extra_instructions` NoxusTipTapDocumentOutput[], nullable: Provide specific rules, goals, or context to guide the co-worker's behavior. - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - `type` string, required - `attrs` object, nullable - `content` NodeOutput[], nullable - `marks` Mark[], nullable - … - `text` string, nullable - `tools` union[] - union - WebResearchSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'web_research' - NoxusQaSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'noxus_qa' - KnowledgeBaseSelectorSettingsOutput: Tool that allows the user to select a knowledge base to answer questions about - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'kb_selector' - KnowledgeBaseQaSettingsOutput: Tool that allows the user to answer questions about a specific pre-selected knowledge base - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'kb_qa' - `kb_id` string, uuid, required - `advanced_search` boolean - `clamp_top_k` boolean - `top_k_window` integer[] - `private` boolean - `enable_browse_tools` boolean - `include_tree_in_instructions` boolean - `tree_depth_in_instructions` integer - `link_references_to_noxus_file` boolean - WorkflowSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string: The lucide icon to display for this tool - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'workflow' - `workflow_id` string, uuid, required: The ID of the workflow to run. Comes unselected, and should be selected by the user. - `inputs_descriptions` object: A dictionary of input names to descriptions. - `input_overrides` object - `version_id` string, nullable: The ID of the version of the workflow to run. Comes unselected, and should be selected by the user. - AttachFileSettings - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string: The lucide icon to display for this tool - `tool_display_message_start` string: The message to display to the user when the tool is called. - `tool_display_message_end` string: The message to display to the user when the tool completes. - `extra_instructions` string, nullable - `requires_approval` boolean - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'attach_file' - HumanInTheLoopSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'human_in_the_loop' - `has_timeout` boolean - `timeout` integer, nullable - `handover_message` string - AgentToolSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'agent_tool' - `agent_id` string, uuid, required: The ID of the agent to call. Comes unselected, and should be selected by the user. - `version_id` string, uuid, nullable - `max_clarification_turns` integer: How many times this co-worker can ask follow-up questions before giving a final answer. Set to 0 to require an immediate response with no back-and-forth. - MemorySettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'memory' - ActionToolSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'action' - `icon` string - `bg_color` string - `icon_color` string - `node_name` string, required: The name of the node/action to execute - `tool_config` ActionToolConfig: Configuration for hardcoded values and exposed parameters in action tools. For each config parameter or input connector, you either: 1. Hardcode it (store value in overrides, don't expose to LLM) 2. Expose it to LLM (add to exposed list, LLM provides value) - … - ExcelSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'excel' - `enable_python_sandbox` boolean - `enable_formula_eval` boolean - `experimental` boolean - DocxSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'docx' - `author_name` string: Author shown on track-change revisions and comments. - FileSystemSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'filesystem' - FileSettingsOutput: Unified file capability for the agent. One master toggle that, when on, lets the user attach files and lets the agent generate + present them (``create_file``, ``present_file``, ``sandbox_publish_file``). Word (.docx) and Excel editing are offered on-demand whenever a matching file is in the session — no separate toggles. The always-on session-file reader (``list_files``/``read_file``) is a separate, hidden capability (``FileSystemSettings``) and is unaffected. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'files' - `allow_html_graphs` boolean: Instruct the agent to use self-contained HTML (Plotly, Chart.js, ...) when generating charts and visualizations. - `show_inline` boolean: Render HTML and images directly in the chat instead of as a card. - `custom_author_name` boolean: Use a custom author for track-change revisions and comments. When off, the agent's name is used. - `author_name` string: Author shown on track-change revisions and comments. Defaults to the agent's name when left unset. - `enable_python_sandbox` boolean: Run pandas/Python over the workbook in the sandbox. - `enable_formula_eval` boolean: Evaluate spreadsheet formulas via the formulas library. - `experimental` boolean: Enable every experimental spreadsheet tool above. - TodoSettingsOutput: Settings for the task tracking toolset. Wraps pydantic-ai-todo's create_todo_toolset() with Noxus settings. Todos are persisted in conversation.memory["todos"]. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'todos' - `enable_subtasks` boolean: Enable subtask hierarchy and dependency features - SandboxSettingsOutput: Settings shared by every version of the sandbox capability: file operations and optional shell execution in an isolated sandbox. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'sandbox' - `persistent` boolean: Keep sandbox running permanently (no idle timeout) - `include_execute` boolean: Include shell command execution tool - `require_write_approval` boolean: Require user approval for file write/edit operations - `require_execute_approval` boolean: Require user approval for shell command execution - `execution_slots` integer: CPU/memory budget of the sandbox in manager slots (1/8 CPU + 256 MiB each) - ScheduleToolSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'schedule_tool' - SubAgentSettingsOutput: Settings for the subagent task delegation toolset. The roster lives on the agent settings (``subagents``); this variant only carries the tool's chat display. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'subagent' - `subagents` SubAgentDefinition[] - … - CodeExecutionSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'code_execution' - `timeout` integer: Maximum execution time in seconds - AgentMemorySettingsOutput: Settings for the persistent agent memory toolset. Memories are stored in assistant.definition["memories"] JSONB and persist across all conversations with this agent. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'agent_memory' - CreateFilesSettingsOutput: Unified setting for the file creation + presentation tools. Enables ``create_file`` (generate text/markdown/html/json/csv/code into the session) and ``present_file`` (surface a session file as a preview/download card in the chat). They share one toggle because they're useless apart — creating a file the user can't see is dead weight, and presenting nothing is a no-op. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'create_files' - `allow_html_graphs` boolean: Instruct the agent to use self-contained HTML (Plotly, Chart.js, ...) when generating charts and visualizations. - `show_inline` boolean: Render HTML and images directly in the chat instead of as a card. - DocumentationSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'documentation' - ImageGenerationSettingsOutput: Image generation quick action. When on, the agent can generate images from text prompts with any image-generation-capable model in the registry; results are surfaced inline in the chat. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'image_generation' - FailureProbeSettingsOutput: Makes an agent fail on demand, so tool-failure handling can be driven in a real conversation instead of only in a unit test. Noxus-internal, behind two independent gates. ``internal_only`` + ``settings.shows_internal_test_tools()`` decides whether the tool exists on this deployment at all — never on-prem, never production, and not configurable. ``AGENT_FAILURE_PROBE_ENABLED`` then decides which of our tenants see it, off by default like every other beta tool. The flag is the weaker of the two on purpose: it is a convenience, and an admin flipping it can never reach a customer deployment, because the deployment gate is checked independently at build time. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'failure_probe' - CodeModeSettingsOutput: Route selected tools through a Monty ``run_code`` sandbox. The tools each capability declared read-only leave the model's native tool list and become async functions inside one ``run_code`` tool: the model writes the orchestration (loops, filters, chained calls) instead of round-tripping every intermediate result through its context. Tool bodies still execute host-side. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'code_mode' - OverflowSettingsOutput: Where an oversized tool output is stored, owned by the quick action. It lives here rather than on ``AgentSettings`` so the capability that reads it carries it: a quick action is only present on agents that added it, so an agent without overflow is never shown knobs for it, and no settings version has to exist to say so. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'overflow' - `storage_backend` 'blob' | 'redis': Where a stored tool output lives until it is read back. - GuardrailsSettingsOutput: Every safety option of an agent on guardrails v2, owned by its quick action. The fields are ``PIIConfig``'s, so v1 agents, which keep them on the agent's own settings, and v2 agents read the same shape. The toggle is the master switch: off builds no shields at all. - `pii_banner` string: Information about PII anonymization - `pii_info` string: Information about PII anonymization - `pii_lock` boolean: Information about the guardrails - `pii_anonimization` object[]: Preferred anonymization method for PII values. - … - `divider1` string - `guardrail_info` string: Information about the guardrails - `guardrail_lock` boolean: Locks the guardrails settings. - `guardrail_latency_banner` string - `violence_and_hate` GuardrailValue - … - `sexual_content` GuardrailValue - … - `guns_and_weapons` GuardrailValue - … - `suicide` GuardrailValue - … - `crime` GuardrailValue - … - `toxicity` GuardrailValue - … - `bias` GuardrailValue - … - `profanity` GuardrailValue - … - `guardrails_scan_output` boolean: Scan the model's response with the same semantic guardrails applied to input. Catches sensitive content sourced from Knowledge Base documents or tools that the input scan cannot see. - `guardrails_output_mode` 'post_stream' | 'pre_stream': How output guardrails are enforced. 'post_stream' streams the response in real-time and warns/clears after the fact. 'pre_stream' buffers the full response, validates it against guardrails, and only emits it if it passes. - `divider2` string - `guardrail_action` union: Choose the action to take when a guardrail is triggered. - … - `guardrail_action_replace` string - `guardrail_action_human` string - `block_after_violations` integer: Block the chat after this many turns with guardrail violations. Set to 0 to never block (per-turn warnings still surface). - `block_message` string: Message shown to the user when the chat is locked. - `divider3` string - `prompt_defense` string - `prompt_lock` boolean: Locks the prompt defense settings - `jailbreak` boolean: Prevents attempts to manipulate or "jailbreak" the model into ignoring safety rules or system instructions (e.g., prompts like "ignore previous instructions"). - `prompt_injection` boolean: Detects and blocks strategies where users attempt to get around restrictions indirectly (e.g., asking the model to 'pretend' or speak as a fictional character to evade policies). - `context_leak` boolean: Prevents the model from unintentionally revealing internal context, instructions, or system messages that are not meant to be shared with the user. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string: The message to display to the user when the tool is called. - `tool_display_message_end` string: The message to display to the user when the tool completes. - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'guardrails' - `settings_display` 'dialog' - `guardrail_engine` 'llm' | 'system_one': How the semantic categories are scanned. The language model makes one call per enabled category. The System 1 classifier answers every category in a single call, faster and far cheaper, and needs a System 1 model enabled under Settings > Models; without one the language model is used. - `workflows_parallel_enabled` boolean: Let the agent start flows without blocking on them, then poll for the result. Off keeps every flow call blocking. - `mcp_servers` MCPServerBinding[]: Workspace MCP credentials this agent can call tools on. - `credential_id` string, uuid, required - `enabled` boolean - `selected_tools` string[], nullable - `convert_unsupported_files_to_text` boolean: When an attached file's type can't be sent to the model as a native attachment (e.g. JSON, CSV, XML, Markdown), inline its text content into the prompt instead of failing the run. PDFs and images are always sent as attachments. - `skills` SkillConfig[] - `name` string, required - `usage_conditions` string - `tags` string[] - `version` string - `author` string - `instructions` string - `enabled` boolean - `resources` object - `subagents_enabled` boolean: Let the agent hand self-contained tasks to a built-in general-purpose subagent, which runs in its own fresh context on the Cost-Effective Models preset. Off keeps every task in the agent's own context. - `subagents` SubAgentDefinition[] - `name` string, required - `usage_conditions` string, required - `instructions` string - `model` string[] - `enabled` boolean - `guidelines_enabled` boolean: Enable behavioral guidelines that shape how the agent responds. - `guidelines` GuidelineConfig[] - `id` string, required - `condition` string, required - `action` string, required - `priority` string - `tags` string[] - `enabled` boolean - `tool_association` ToolAssociationConfig: Tool enforcement configuration for a guideline. - `tool_names` string[] - `mode` string - `tool_associations` ToolAssociationConfig[] - `tool_names` string[] - `mode` string - `glossary` GlossaryTermConfig[] - `term` string, required - `definition` string, required - `synonyms` string[] - `memories` AgentMemoryEntry[]: Persistent memory across all conversations - `id` string, required - `content` string, required - `category` string - `created_at` string - `updated_at` string - `hidden_tools` string[]: Tool calls that never appear in the chat activity feed. - `deployment_custom_instructions` string, nullable - `variables_enabled` boolean: Create external values that can be mapped to flow inputs. The agent receives these external values through its widget deployment. - `variables` VariableDefinition[] - `name` string, required - `workflows` WorkflowSettingsOutput[]: Workflows the agent can run - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string: The lucide icon to display for this tool - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'workflow' - `workflow_id` string, uuid, required: The ID of the workflow to run. Comes unselected, and should be selected by the user. - `inputs_descriptions` object: A dictionary of input names to descriptions. - `input_overrides` object - `version_id` string, nullable: The ID of the version of the workflow to run. Comes unselected, and should be selected by the user. - `knowledge_bases` KnowledgeBaseQaSettingsOutput[]: Knowledge bases the agent can search - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'kb_qa' - `kb_id` string, uuid, required - `advanced_search` boolean - `clamp_top_k` boolean - `top_k_window` integer[] - `private` boolean - `enable_browse_tools` boolean - `include_tree_in_instructions` boolean - `tree_depth_in_instructions` integer - `link_references_to_noxus_file` boolean - `multi_kb_search` boolean: Let one search cover several knowledge bases at once, merging their results. Off means every search targets a single one. - `citation_instructions_enabled` boolean: Write your own rules for what the co-worker cites from its knowledge bases. Off keeps the built-in citation behaviour. - `citation_instructions` NoxusTipTapDocumentOutput[], nullable: Tell the co-worker what to cite from its knowledge bases, when a claim needs a source, and how much to quote. The reference tag format itself is fixed and always applies. - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - `type` string, required - `attrs` object, nullable - `content` NodeOutput[], nullable - `marks` Mark[], nullable - … - `text` string, nullable - `agent_files` AgentFileConfigOutput[]: Files pre-loaded into every chat. Each is added to the session (readable via the file tools) with its context shown to the agent. - `file` File, required - `id` string, nullable - `uri` string, required - `name` string, required - `content_type` string - `size` integer, nullable - `content_metadata` union - PdfMetadata: Structural metadata for a PDF file. ``page_labels`` is only set when the document's labels differ from plain 1..N numbering (cover pages, roman numerals), mapping physical page index to label. - … - SpreadsheetMetadata: Structural metadata for a spreadsheet (Excel) file. - … - TabularMetadata: Structural metadata for a delimited-text (CSV/TSV) file. - … - ImageMetadata: Structural metadata for an image file. - … - PresentationMetadata: Structural metadata for a presentation (PowerPoint) file. - … - DocumentMetadata: Structural metadata for a word-processing document (Word). ``page_count`` is the authoring app's cached count from ``docProps/app.xml`` — present only for Word-authored files, ``None`` when the document came from another tool (Google Docs, LibreOffice, python-docx) that does not write the cache. ``word_count`` / ``paragraph_count`` are derived from the body text and always populated. - … - `source_type` 'Document' | 'Google Drive' | 'Notion' | 'Website' | 'OneDrive' | 'Slack' | 'Linear' | 'Github' | 'Teams' | 'Sharepoint' | 'ServiceNow' | 'Custom' - `source_metadata` union - SourceMetadata: Base class for source metadata across different source types. ``external_id`` is the source's stable id for this item (Drive file id, OneDrive item id, URL for websites, etc.). Datasource ``download_file`` implementations stamp it explicitly, and the DB materialisation trigger projects it onto ``KnowledgeBaseDocument.m_source_external_id`` — which is the join key the KB sync engine uses to find existing docs by their source id. - … - object - `context` NoxusTipTapDocumentOutput - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `enabled` boolean - `delimiter` string, nullable - `databases` AgentDatabaseConfigOutput[]: SQL databases the agent can query read-only. Each references a database credential; access is deferred, so the agent connects only when a query is needed. - `credential_id` string, uuid, required - `context` NoxusTipTapDocumentOutput - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `enabled` boolean - `selected_tables` string[], nullable - `max_rows` integer - `query_timeout_seconds` integer - `tables` AgentTableConfigOutput[]: The workspace tables this agent can use. Add context for each so the agent knows what it holds and when to use it. - `table_id` string, uuid, required - `context` NoxusTipTapDocumentOutput - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `enabled` boolean - `read_only` boolean - `max_rows` integer - `query_timeout_seconds` integer - `coworkers` AgentToolSettingsOutput[]: Co-workers the agent can call - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'agent_tool' - `agent_id` string, uuid, required: The ID of the agent to call. Comes unselected, and should be selected by the user. - `version_id` string, uuid, nullable - `max_clarification_turns` integer: How many times this co-worker can ask follow-up questions before giving a final answer. Set to 0 to require an immediate response with no back-and-forth. - `base_tools` union[]: Quick action tools the agent can use - union - WebResearchSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'web_research' - HumanInTheLoopSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'human_in_the_loop' - `has_timeout` boolean - `timeout` integer, nullable - `handover_message` string - MemorySettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'memory' - AttachFileSettings - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string: The lucide icon to display for this tool - `tool_display_message_start` string: The message to display to the user when the tool is called. - `tool_display_message_end` string: The message to display to the user when the tool completes. - `extra_instructions` string, nullable - `requires_approval` boolean - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'attach_file' - FileSystemSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'filesystem' - FileSettingsOutput: Unified file capability for the agent. One master toggle that, when on, lets the user attach files and lets the agent generate + present them (``create_file``, ``present_file``, ``sandbox_publish_file``). Word (.docx) and Excel editing are offered on-demand whenever a matching file is in the session — no separate toggles. The always-on session-file reader (``list_files``/``read_file``) is a separate, hidden capability (``FileSystemSettings``) and is unaffected. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'files' - `allow_html_graphs` boolean: Instruct the agent to use self-contained HTML (Plotly, Chart.js, ...) when generating charts and visualizations. - `show_inline` boolean: Render HTML and images directly in the chat instead of as a card. - `custom_author_name` boolean: Use a custom author for track-change revisions and comments. When off, the agent's name is used. - `author_name` string: Author shown on track-change revisions and comments. Defaults to the agent's name when left unset. - `enable_python_sandbox` boolean: Run pandas/Python over the workbook in the sandbox. - `enable_formula_eval` boolean: Evaluate spreadsheet formulas via the formulas library. - `experimental` boolean: Enable every experimental spreadsheet tool above. - CodeExecutionSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'code_execution' - `timeout` integer: Maximum execution time in seconds - ExcelSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'excel' - `enable_python_sandbox` boolean - `enable_formula_eval` boolean - `experimental` boolean - DocxSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'docx' - `author_name` string: Author shown on track-change revisions and comments. - NoxusQaSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'noxus_qa' - KnowledgeBaseSelectorSettingsOutput: Tool that allows the user to select a knowledge base to answer questions about - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean: Whether the tool is enabled - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'kb_selector' - TodoSettingsOutput: Settings for the task tracking toolset. Wraps pydantic-ai-todo's create_todo_toolset() with Noxus settings. Todos are persisted in conversation.memory["todos"]. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'todos' - `enable_subtasks` boolean: Enable subtask hierarchy and dependency features - SandboxSettingsOutput: Settings shared by every version of the sandbox capability: file operations and optional shell execution in an isolated sandbox. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'sandbox' - `persistent` boolean: Keep sandbox running permanently (no idle timeout) - `include_execute` boolean: Include shell command execution tool - `require_write_approval` boolean: Require user approval for file write/edit operations - `require_execute_approval` boolean: Require user approval for shell command execution - `execution_slots` integer: CPU/memory budget of the sandbox in manager slots (1/8 CPU + 256 MiB each) - ScheduleToolSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'schedule_tool' - SubAgentSettingsOutput: Settings for the subagent task delegation toolset. The roster lives on the agent settings (``subagents``); this variant only carries the tool's chat display. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'subagent' - `subagents` SubAgentDefinition[] - … - AgentMemorySettingsOutput: Settings for the persistent agent memory toolset. Memories are stored in assistant.definition["memories"] JSONB and persist across all conversations with this agent. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'agent_memory' - CreateFilesSettingsOutput: Unified setting for the file creation + presentation tools. Enables ``create_file`` (generate text/markdown/html/json/csv/code into the session) and ``present_file`` (surface a session file as a preview/download card in the chat). They share one toggle because they're useless apart — creating a file the user can't see is dead weight, and presenting nothing is a no-op. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean - `type` 'create_files' - `allow_html_graphs` boolean: Instruct the agent to use self-contained HTML (Plotly, Chart.js, ...) when generating charts and visualizations. - `show_inline` boolean: Render HTML and images directly in the chat instead of as a card. - DocumentationSettingsOutput - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'documentation' - ImageGenerationSettingsOutput: Image generation quick action. When on, the agent can generate images from text prompts with any image-generation-capable model in the registry; results are surfaced inline in the chat. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'image_generation' - FailureProbeSettingsOutput: Makes an agent fail on demand, so tool-failure handling can be driven in a real conversation instead of only in a unit test. Noxus-internal, behind two independent gates. ``internal_only`` + ``settings.shows_internal_test_tools()`` decides whether the tool exists on this deployment at all — never on-prem, never production, and not configurable. ``AGENT_FAILURE_PROBE_ENABLED`` then decides which of our tenants see it, off by default like every other beta tool. The flag is the weaker of the two on purpose: it is a convenience, and an admin flipping it can never reach a customer deployment, because the deployment gate is checked independently at build time. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'failure_probe' - CodeModeSettingsOutput: Route selected tools through a Monty ``run_code`` sandbox. The tools each capability declared read-only leave the model's native tool list and become async functions inside one ``run_code`` tool: the model writes the orchestration (loops, filters, chained calls) instead of round-tripping every intermediate result through its context. Tool bodies still execute host-side. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'code_mode' - OverflowSettingsOutput: Where an oversized tool output is stored, owned by the quick action. It lives here rather than on ``AgentSettings`` so the capability that reads it carries it: a quick action is only present on agents that added it, so an agent without overflow is never shown knobs for it, and no settings version has to exist to say so. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'overflow' - `storage_backend` 'blob' | 'redis': Where a stored tool output lives until it is read back. - GuardrailsSettingsOutput: Every safety option of an agent on guardrails v2, owned by its quick action. The fields are ``PIIConfig``'s, so v1 agents, which keep them on the agent's own settings, and v2 agents read the same shape. The toggle is the master switch: off builds no shields at all. - `pii_banner` string: Information about PII anonymization - `pii_info` string: Information about PII anonymization - `pii_lock` boolean: Information about the guardrails - `pii_anonimization` object[]: Preferred anonymization method for PII values. - … - `divider1` string - `guardrail_info` string: Information about the guardrails - `guardrail_lock` boolean: Locks the guardrails settings. - `guardrail_latency_banner` string - `violence_and_hate` GuardrailValue - … - `sexual_content` GuardrailValue - … - `guns_and_weapons` GuardrailValue - … - `suicide` GuardrailValue - … - `crime` GuardrailValue - … - `toxicity` GuardrailValue - … - `bias` GuardrailValue - … - `profanity` GuardrailValue - … - `guardrails_scan_output` boolean: Scan the model's response with the same semantic guardrails applied to input. Catches sensitive content sourced from Knowledge Base documents or tools that the input scan cannot see. - `guardrails_output_mode` 'post_stream' | 'pre_stream': How output guardrails are enforced. 'post_stream' streams the response in real-time and warns/clears after the fact. 'pre_stream' buffers the full response, validates it against guardrails, and only emits it if it passes. - `divider2` string - `guardrail_action` union: Choose the action to take when a guardrail is triggered. - … - `guardrail_action_replace` string - `guardrail_action_human` string - `block_after_violations` integer: Block the chat after this many turns with guardrail violations. Set to 0 to never block (per-turn warnings still surface). - `block_message` string: Message shown to the user when the chat is locked. - `divider3` string - `prompt_defense` string - `prompt_lock` boolean: Locks the prompt defense settings - `jailbreak` boolean: Prevents attempts to manipulate or "jailbreak" the model into ignoring safety rules or system instructions (e.g., prompts like "ignore previous instructions"). - `prompt_injection` boolean: Detects and blocks strategies where users attempt to get around restrictions indirectly (e.g., asking the model to 'pretend' or speak as a fictional character to evade policies). - `context_leak` boolean: Prevents the model from unintentionally revealing internal context, instructions, or system messages that are not meant to be shared with the user. - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents': The category of the tool - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string: The message to display to the user when the tool is called. - `tool_display_message_end` string: The message to display to the user when the tool completes. - `extra_instructions` object, nullable: Any extra instructions for the tool - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `icon` string - `bg_color` string - `icon_color` string - `deprecated` boolean: Whether this tool type is deprecated and should be hidden from the UI - `type` 'guardrails' - `settings_display` 'dialog' - `guardrail_engine` 'llm' | 'system_one': How the semantic categories are scanned. The language model makes one call per enabled category. The System 1 classifier answers every category in a single call, faster and far cheaper, and needs a System 1 model enabled under Settings > Models; without one the language model is used. - `actions` ActionToolSettingsOutput[]: Action nodes the agent can execute - `title` string - `description` string - `category` 'QuickAction' | 'Flows' | 'Knowledge' | 'Actions' | 'MCP' | 'Agents' - `enabled` boolean - `tool_display_icon` string - `tool_display_message_start` string - `tool_display_message_end` string - `extra_instructions` object, nullable: Any extra instructions for the tool - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - `requires_approval` boolean: Whether tool calls require user approval before execution - `visible` boolean: Whether the tool is visible - `type` 'action' - `icon` string - `bg_color` string - `icon_color` string - `node_name` string, required: The name of the node/action to execute - `tool_config` ActionToolConfig: Configuration for hardcoded values and exposed parameters in action tools. For each config parameter or input connector, you either: 1. Hardcode it (store value in overrides, don't expose to LLM) 2. Expose it to LLM (add to exposed list, LLM provides value) - `config_overrides` object - `input_overrides` object - `exposed_configs` string[] - `exposed_inputs` string[] - `insights_enabled` boolean: Enable conversation analysis and insights dashboard. - `insights_csat_enabled` boolean: Collect customer satisfaction (CSAT) ratings at end of conversations. - `insights_csat_sampling_rate` number: Fraction of conversations that get a CSAT prompt (1.0 = every conversation). - `insights_csat_agent_tool_enabled` boolean: Let the agent decide when to show the CSAT form via a tool call. - `insights_csat_agent_tool_instructions` union: Custom instructions for when the agent should request customer feedback. - object - `text` string, required - `definition` TiptapDocumentOutput - `type` string - `content` NodeOutput[] - … - string - `insights_csat_auto_enabled` boolean: Automatically show the CSAT form at the end of an agent turn (rule-based, no tool call). Independent of the agent feedback tool. - `insights_csat_auto_min_messages` integer: Only auto-request feedback once the user has sent at least this many messages. - `insights_custom_definitions` CustomInsightDefinition[]: User-defined situations to detect in each conversation. - `name` string, required - `prompt` string, required - `scope` 'all' | 'resolved' | 'unresolved' | 'escalated' | 'abandoned' | 'rated-high' | 'rated-low' - `tool_grant_scope` 'workspace' | 'user': Whether 'Always Allow' tool grants apply to all workspace users or only the granting user - `assistant_id` string, uuid, nullable - `status` 'queued' | 'running' | 'failed' | 'completed' | 'stopped' | 'awaiting_human_feedback', required - `manual` boolean - `preview` boolean - `assistant_version_id` string, uuid, nullable - `chat_transport` string, nullable - `channel_ref` object, nullable - `deployment_id` string, uuid, nullable - `public_access_token` string, nullable - `todos` object[] - `is_locked` boolean - `lock_message` string, nullable - `total` integer, required - `page` integer, required - `size` integer, required - `pages` integer, required ## Other responses - `422`: Validation Error --- [API](https://skmtc.dev/noxus/apis/noxus-backend.md) - [All operations](https://skmtc.dev/noxus/apis/noxus-backend/llms.txt) - [OpenAPI document](https://skmtc.dev/noxus/apis/noxus-backend/revisions/0e37810b7ad5?raw)