EOS

External Agent Triggers

Adonis EOS allows triggering AI agents from external sources such as webhooks, CI/CD pipelines, or third-party automation tools (like n8n, Make, or custom scripts).

There are two primary ways to trigger agents externally:

  1. MCP (Model Context Protocol): Best for AI-to-AI communication.

  2. REST API: Best for simple webhook or script triggers.

1. Triggering via MCP (Recommended for AI systems)

The EOS MCP server includes a run_agent tool specifically designed for external triggers.

Tool: run_agent

Parameters:

  • agentId (string, required): The ID of the agent to run (e.g., general_assistant).

  • postId (string, optional): The ID of the post the agent should work on.

  • scope (string, optional): The scope to run in (default: dropdown).

  • context (object, optional): Additional context parameters for the agent.

  • openEndedContext (string, optional): A freeform prompt for the agent (if supported).

Example (n8n MCP Client Tool):

json 
{
  "agentId": "graphic_designer",
  "postId": "uuid-of-post",
  "scope": "dropdown",
  "context": {
    "viewMode": "source"
  }
}

2. Triggering via REST API

If you prefer a standard HTTP request, you can use the external trigger endpoint.

Endpoint: POST /api/public/agents/:agentId/run

Authentication: Requires a secret key sent in one of the following headers:

  • X-Agent-Key: YOUR_SECRET

  • Authorization: Bearer YOUR_SECRET

The secret is determined by the AGENT_TRIGGER_SECRET environment variable. If not set, it falls back to MCP_AUTH_TOKEN.

Request Body:

json 
{
  "postId": "uuid-of-post",
  "scope": "dropdown",
  "context": {
    "any": "value"
  },
  "openEndedContext": "Summarize this post"
}

cURL Example:

bash 
curl -X POST http://your-eos-site.com/api/public/agents/general_assistant/run \
  -H "X-Agent-Key: your-secret-token" \
  -H "Content-Type: application/json" \
  -d '{
    "postId": "550e8400-e29b-41d4-a716-446655440000",
    "openEndedContext": "Check for spelling errors"
  }'

Security Considerations

  • Secret Keys: Always set a strong AGENT_TRIGGER_SECRET in your .env file.

  • Least Privilege: Agents triggered externally still follow the same safety rules as internal agents. Suggestions are typically staged to AI Review rather than applied directly to the live post, unless the agent is specifically configured otherwise.

  • Rate Limiting: Public endpoints are subject to EOS global rate limits. For high-volume triggers, ensure your IP is allowlisted or adjust config/cms.ts rate limit settings.