Agent Integration

How AI agents can consume TrigRun documentation and operate the platform programmatically.

TrigRun is designed for AI agents to use directly. Its public API is machine-readable end to end, and the product MCP server covers the most common operational workflows around request-based jobs, executions, secrets, and notifications.

Quick access

Resource	URL	Format	Use case
OpenAPI spec	`https://api.trigrun.com/openapi.json`	OpenAPI 3.1 JSON	Full API contract — request/response schemas, auth, examples
llms.txt	`https://api.trigrun.com/llms.txt`	Markdown	Concise API summary for LLM context windows
llms-full.txt	`https://api.trigrun.com/llms-full.txt`	Markdown	Complete documentation in one file
MCP server	See Product MCP	MCP protocol	Operate common TrigRun workflows from Claude, Cursor, or any MCP client

For coding agents

If you're an AI agent writing code that integrates with TrigRun:

Fetch /openapi.json — This is your primary reference. It contains every endpoint, request/response schema, auth requirements, and examples.
Auth header format: Authorization: Bearer <token> — either a JWT from login or an API token starting with cron_pat_.
Error format: All errors return { "error": { "code": "...", "message": "...", "details": ... } }.
Base URL: https://api.trigrun.com

Minimal example: create a job

curl -X POST https://api.trigrun.com/v1/jobs \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "agent-created-job",
    "kind": "cron",
    "schedule": { "cron": "*/5 * * * *" },
    "request": {
      "url": "https://httpbin.org/post",
      "method": "POST"
    }
  }'

Common agent workflows

Task	Endpoint	Method
Create a job	`/v1/jobs`	POST
List all jobs	`/v1/jobs`	GET
Trigger a job now	`/v1/jobs/:id/run-now`	POST
Check execution result	`/v1/executions/:id`	GET
List failed executions	`/v1/executions?status=failed`	GET
Replay a failed execution	`/v1/executions/:id/replay`	POST
Create a secret	`/v1/secrets`	POST
Import from crontab	`/v1/jobs/import`	POST

For MCP-native agents

If you're an agent with MCP support (Claude Desktop, Claude Code, Cursor):

Connect to the TrigRun MCP server — see Product MCP setup
Use tools like list_jobs, create_job, get_execution directly
Every tool requires a token argument (API token or JWT)

For features not yet exposed as MCP tools, use the REST API directly. Today that includes monitors and step-driven workflow/webhook-trigger job creation.

For context-loading agents

If you need to load TrigRun docs into your context:

Small context: Fetch https://api.trigrun.com/llms.txt (~2K tokens) for a concise API overview
Full context: Fetch https://api.trigrun.com/llms-full.txt (~15K tokens) for complete docs with examples
Specific page: Use the docs navigation or search, then fetch the page's raw Markdown

Idempotency

When creating jobs programmatically, use the Idempotency-Key header to prevent duplicates:

curl -X POST https://api.trigrun.com/v1/jobs \
  -H "Idempotency-Key: deploy-v2.1.0" \
  ...

If a job with the same key exists, the existing job is returned.