MCPBundler Docs
MCPBundler Docs
Back to marketplaceGetting StartedAuthenticationTutorial - Bundle to Connected Agent
API Reference
Agent A2A Tools
Agent Channels
Agent Configuration (Staging)
Create Agent Custom ToolCreate Agent GuardrailRemove Agent System ToolDelete Agent Custom ToolDelete Agent GuardrailDiscard Staged ChangesDiff Stage Against Latest PublishedGet Stage WorkflowList Agent System ToolsList Agent Custom ToolsList Agent GuardrailsUpdate Staged Voice SettingsUpdate Staged Agent SettingsPatch Stage WorkflowRegenerate Widget Embed KeyUpdate Agent Custom ToolEnable or Configure Agent System Tool
Agent Configuration (Versions)
Agent Tools
Agents
Auth
Billing
Bundler
Bundler Tool
Bundles
Call Campaigns
Catalog
Conversations
Credentials
Deployments
Discover
Embedding
Fleet
Knowledge Bases
LLM
MCP Listings
Media
OAuth Clients
OAuth2
Observability
Organisations
Realtime
Reviews
Roadmap
SCIM
STT
Suppression
Triggers
TTS
Users
Utils
Voices
Webhooks
Workflows
API ReferenceAgent Configuration (Staging)

Create Agent Guardrail

POST
/v1/agents/{agent_id}/stage/guardrails

Reflects unpublished staged state - see GET /agents/{agent_id}/versions/latest for what's actually live.

Authorization

OAuth2AuthorizationCodeBearer
AuthorizationBearer <token>

In: header

Path Parameters

agent_id*Agent Id
Formatuuid

Query Parameters

verify_locally?Verify Locally
Defaulttrue

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

Response Body

application/json

application/json

curl -X POST "https://example.com/v1/agents/497f6eca-6276-4993-bfeb-53cbbbba6f08/stage/guardrails" \  -H "Content-Type: application/json" \  -d '{    "hook_point": "before_llm_call",    "webhook_url": "http://example.com",    "webhook_secret": "string"  }'
{  "id": "497f6eca-6276-4993-bfeb-53cbbbba6f08",  "hook_point": "string",  "webhook_url": "string",  "fail_open": true,  "timeout_ms": 0,  "enabled": true,  "created_at": "2019-08-24T14:15:22Z",  "updated_at": "2019-08-24T14:15:22Z"}
{  "detail": [    {      "loc": [        "string"      ],      "msg": "string",      "type": "string"    }  ]}

Create Agent Custom Tool

Previous Page

Remove Agent System Tool

Next Page

hook_point*Hook Point

Only 'before_tool_call' and 'after_tool_call' guardrails are currently enforced by agent-runtime. A guardrail configured for 'before_llm_call' or 'after_llm_response' is accepted and stored but is not currently invoked during a conversation - it has no effect until those hook points are wired up.

Value in

  • "before_llm_call"
  • "before_tool_call"
  • "after_tool_call"
  • "after_llm_response"
webhook_url*Webhook Url
Formaturi
Length1 <= length <= 2083
webhook_secret*Webhook Secret
Length1 <= length
fail_open?Fail Open
Defaultfalse
timeout_ms?Timeout Ms

Reserved for a future per-call webhook timeout. The effective timeout ceiling is currently fixed at approximately 3s connect / 5s read regardless of the value configured here - agent-runtime does not yet thread this value through to the underlying webhook client.

Range500 <= value <= 15000
Default5000
enabled?Enabled
Defaulttrue