A2A Agent Card

The /.well-known/agent-card.json endpoint implements the Google A2A (Agent-to-Agent) discovery protocol. External agents and peer platforms query this URL to discover what capabilities Rensei exposes and how to authenticate dispatch requests.

Discovery endpoint

GET /.well-known/agent-card.json

No authentication required. This endpoint is fully public per the A2A spec.

Response headers

Header	Value
`Content-Type`	`application/json`
`Cache-Control`	`no-store`
`Access-Control-Allow-Origin`	`*`

Two modes of operation

The endpoint supports two distinct modes depending on whether an ?org query parameter is provided.

Platform-level card (no query param)

Returns the Rensei platform's own agent card, aggregating all scope='system' agent card rows. This is what peer platforms use to discover the platform's top-level capabilities.

curl https://app.rensei.ai/.well-known/agent-card.json

Example response

{
  "@type": "AgentCard",
  "name": "Rensei Platform",
  "description": "AI-native software development lifecycle orchestration.",
  "url": "https://app.rensei.ai",
  "version": "1.0.0",
  "capabilities": {
    "streaming": false,
    "pushNotifications": false
  },
  "skills": [
    {
      "id": "sdlc.execute",
      "name": "Execute SDLC workflow",
      "description": "Run a full software development lifecycle workflow for a given issue.",
      "inputModes": ["text"],
      "outputModes": ["text", "artifacts"]
    }
  ],
  "authentication": {
    "schemes": ["Bearer"]
  }
}

Per-org card (with `?org=<slug>`)

Returns the agent card for a specific organization, aggregating scope='org' agent card rows where trust.publish_to_a2a=true. Use this to discover org-specific agent skills and trust policies.

curl "https://app.rensei.ai/.well-known/agent-card.json?org=my-org"

If no publishable agents exist for the org, the endpoint returns 404 Not Found.

Only agents with trust.publish_to_a2a = true in their agent card configuration appear in the per-org discovery response. You can control this in Settings > Agent Cards or via the API.

Example response

{
  "@type": "AgentCard",
  "name": "My Org Agent Surface",
  "url": "https://app.rensei.ai",
  "skills": [
    {
      "id": "my-org.code-review",
      "name": "Code review agent",
      "description": "Reviews pull requests and posts structured feedback.",
      "tags": ["code-review", "github"]
    }
  ]
}

A2A agent card schema

Rensei publishes agent cards conforming to the Google A2A specification. Key fields:

Prop

Type

Authoring agent cards

Agent cards are authored in YAML and managed via Settings > Agent Cards or the agent-cards API. A card that publishes to A2A requires trust.publish_to_a2a: true:

name: "Code Review Agent"
description: "Reviews PRs and posts structured feedback to Linear."
scope: org
trust:
  publish_to_a2a: true
skills:
  - id: code-review
    name: "Review a pull request"
    description: "Analyzes code changes and posts a structured review."
    inputModes:
      - text
    outputModes:
      - text

See Agent Cards overview for the full YAML schema, scope cascade, and partials system.

Audit keys discovery

For audit verification, a companion endpoint publishes the organization's JWKS (signing keys for audit events):

curl "https://app.rensei.ai/.well-known/audit-keys/<workspace_id>"

See Audit JWKS for details.

A2A Dispatch - listing dispatches and managing agent registrations
Agent Cards overview - authoring agent cards, scope model
A2A Policies - Cedar-based trust policies for A2A dispatch
MCP Session Tools - A2A agents exposed as MCP tools

On this page