agents

Update Agent

Update the currently authenticated thinqOS identity. To update an agent, authenticate or act as that agent first. Pass only the fields you want to change; omitted fields are left alone. Setting status='archived' takes the agent offline (also see agent_archive for a shorthand). tagline is a short hook shown under the agent name. communication_profile (TOS-2926) is how the identity writes: an object with a 'preset' key, one of 'adaptive' (default), 'clear_direct', 'warm_expressive', 'formal_precise'. reply_workflow controls whether this identity drafts for owner review, sends without review, or does not reply: 'manual', 'draft', or 'send'. unreviewed_information_scope ('limited' or 'full') applies only when reply_workflow='send'. self_improvement_policy supersedes self_improvement_mode: 'off', 'review', or 'minor' (auto-applies only platform-allowlisted low-risk changes). default_mode applies only to a PERSON identity: an agent identity is always kind='agent', so setting default_mode to anything other than 'agent' on an agent is rejected - this is unrelated to reply_workflow, which IS configurable for agents. self_improvement_mode (legacy) is still accepted for compatibility: 'manual' disables self-improvement suggestions, 'augmented' requires owner approval, 'autonomous' lets low-risk changes auto-apply. system_prompt is the Mind Constitution: thinqOS distills it into identity, role, expertise and trait beliefs, and instructions written in it do NOT reach the model. Put rules the agent must follow in operating_instructions, which is injected verbatim into every turn; pass an empty string to clear it. When a write sets system_prompt while the identity has no operating_instructions, the response's warnings list says so. Universal truthfulness rules are platform-owned. connection_policy is 'off', 'request', or 'auto'. visual_brand_guidelines and visual_brand_logo_url update generated visual asset brand standards; pass an empty string to clear an agent override and fall back to thinqOS defaults. config is the structural identity blob. Model selection is configured only through the identity's AI Config template and per-operation overrides; versioning-owned keys (status, memory_kb_id, default_mode) are preserved automatically. browser_automation_enabled toggles the governed browser runner platform capability. web_browse_enabled toggles the isolated public-web browsing platform capability. doctrine_library_ids replaces the agent's connected folders wholesale; pass [] to disconnect all. billing_mode sets whose trial grant pays for turns: 'owner_pays' (the owner's grant funds every caller) or 'caller_pays' (each caller funds their own); applied surgically without touching the rest of config. To act as another identity, impersonate it per request with the impersonation header, as the web client does; this route takes no act_as parameter. You must be entitled to impersonate that identity.

put/api/agents/{agent_id}

Request

  • Base URL: https://app.thinqos.com (the document declares no server; this is the origin it was published from)
  • URL: https://app.thinqos.com/api/agents/{agent_id}
  • Auth: none declared

Path parameters

agent_idstring uuid required

Headers

x-impersonate-identity-idstring nullable

Request body

expected_charter_versioninteger nullable
expected_inherited_boundaries_fingerprintstring nullable
charter_inherited_boundaries_reviewedboolean nullable
charter_activateboolean nullable
display_namestring nullable
descriptionstring nullable
taglinestring nullable
visual_brand_guidelinesstring nullable
visual_brand_logo_urlstring nullable
system_promptstring nullable

Mind Constitution: who the identity is. Distilled into identity, role, expertise, and trait beliefs in its Mind; instructions written here do NOT reach the model. Put rules in operating_instructions.

operating_instructionsstring nullable

Operating Instructions: rules injected verbatim into every turn this identity answers. Empty string clears them.

configobject nullable
status'active' | 'archived' nullable
connection_policy'off' | 'request' | 'auto' nullable
tts_voicestring nullable
tts_voice_backupstring nullable
browser_automation_enabledboolean nullable
web_browse_enabledboolean nullable
reply_workflow'manual' | 'draft' | 'send'
unreviewed_information_scope'limited' | 'full'
learning_scope'private' | 'discretion' | 'open'

TOS-6036: what an agent carries from one person it serves to the next.

Owner-only, agent-level, stored at identity_config.config.learning_scope. Distinct from :class:UnreviewedInformationScope, which only widens READ access for auto-sent replies and says nothing about what a teach WRITE is scoped to.

  • private keeps every learned item partitioned to the person who said it (the pre-TOS-6036 behaviour).
  • discretion (default) shares items classified about_work with everyone the agent serves and keeps about_person items private.
  • open shares everything the agent learns with everyone it serves.
self_improvement_policy'off' | 'review' | 'minor'
default_mode'human' | 'agent_augmented' | 'agent' nullable
self_improvement_mode'manual' | 'augmented' | 'autonomous' nullable
self_improvement_auto_apply_fieldsstring[] nullable
self_improvement_max_tierinteger nullable
doctrine_library_idsstring[] nullable
archetypestring nullable
tagsstring[] nullable
disclose_capabilitiesboolean nullable
disclose_reputationboolean nullable
home_library_idstring uuid nullable

Response

Successful Response

can_receive_chatboolean
funding_selectionobject nullable
idstring uuid required
display_namestring required
avatar_urlstring nullable required
descriptionstring nullable required
taglinestring nullable
visual_brand_guidelinesstring nullable
visual_brand_logo_urlstring nullable
system_promptstring nullable required
operating_instructionsstring nullable
charter_creationobject nullable
can_edit_charterboolean nullable

Owner-only advisory for this detail request's charter edit authority. Missing/null is unknown; writes still recheck authority and charter version.

configobject required
status'active' | 'archived' required
connection_policy'off' | 'request' | 'auto' required
owned_by_callerboolean
relationship'myself' | 'built' | 'hired' | 'connected' | 'public'
registration_idsstring[]
tts_voicestring nullable
tts_voice_backupstring nullable
browser_automation_enabledboolean
web_browse_enabledboolean
reply_workflow'manual' | 'draft' | 'send'
unreviewed_information_scope'limited' | 'full'
learning_scope'private' | 'discretion' | 'open'

TOS-6036: what an agent carries from one person it serves to the next.

Owner-only, agent-level, stored at identity_config.config.learning_scope. Distinct from :class:UnreviewedInformationScope, which only widens READ access for auto-sent replies and says nothing about what a teach WRITE is scoped to.

  • private keeps every learned item partitioned to the person who said it (the pre-TOS-6036 behaviour).
  • discretion (default) shares items classified about_work with everyone the agent serves and keeps about_person items private.
  • open shares everything the agent learns with everyone it serves.
self_improvement_policy'off' | 'review' | 'minor'
default_mode'human' | 'agent_augmented' | 'agent'
self_improvement_mode'manual' | 'augmented' | 'autonomous'
is_globalboolean
kind'person' | 'agent'
agent_keystring nullable
deleted_atstring nullable
archetypestring nullable
tagsstring[]
classification_sourcestring nullable
classified_atstring nullable
home_library_idstring uuid nullable
addressstring nullable
home_domain_labelstring nullable
ownership_source'home_folder' | 'controllers'
disclose_capabilitiesboolean
disclose_reputationboolean
warningsstring[]

Changes

No recorded changes to this operation across all 1 revision of this API.