Create Agent From Config

Input configuration for the agent.

Output configuration for the agent.

Inline LLM configuration embedded on the agent.

Carries provider, credentials, model, and tuning controls as a single object. Values can be copied from a saved AgentLlmConnection (which carries default model, tuning and reasoning controls) and overridden per agent; the object is stored as sent and used by the agent runtime at deploy.

Two credential shapes (exactly one, same rule as ValidateLlmRequest):

1. Inline - apiKey is filled directly. Streamkap encrypts it at rest and decrypts it only at deploy / test-run.
2. Linked - llmConnectionId references one of the tenant's saved LLM connections and apiKey is left empty. Streamkap resolves the real key server-side at deploy and at test-run, so a client never needs to hold or re-send the stored key when editing an existing agent.

provider == ollama accepts an empty key without a link (the runtime is local and unauthenticated).

Prompt configuration for the agent.

The final system prompt is assembled automatically from the agent type template + output schema + tool descriptions + customInstructions. The system field stores the assembled result (set by Streamkap, not the user).

Top-level MCP server config shared across all MCP tools.

Two shapes:

• Streamkap MCP: projectKeyId set; serverUrl and headers are resolved server-side at deploy time from the agentic-enabled Project Key.
• External MCP: serverUrl + headers set inline; projectKeyId omitted. Headers are composed at deploy time from the matching saved external connection (bearer / custom-header).

Agent tools (MCP, HTTP, transform)

Memory configuration for the agent.

Knowledge bases this agent can query at runtime for RAG

Processing configuration for the agent.