Help Center › Agents Developer Help › Developer API

Agent info, models, and capabilities

Availability: All plans. Requires a deployed agent and an agent token.

Three read-only endpoints tell your client everything about the agent behind your token — who it is, what model it runs, and what it can accept. Call them once at startup; together they replace any hardcoded assumptions about the agent.

All three are GET requests authenticated the usual way:

Authorization: Bearer <AGENT_TOKEN>

GET /v3/agent/info — the whole picture

One call that consolidates configuration, model metadata, and capabilities — designed for frontend initialization and debugging.

{
  "agent_id": "1042",
  "configuration": {
    "name": "Support Assistant",
    "goal": "Resolve customer questions about orders and returns.",
    "instructions": "Always confirm the order number before acting…",
    "guardrails": "Never discuss pricing exceptions.",
    "tone": "friendly",
    "tools": ["rag_tool", "order_lookup_http"],
    "environment_variables": ["CRM_BASE_URL"]
  },
  "model": {
    "name": "gpt-4o",
    "display_name": "GPT-4o",
    "provider": "openai",
    "context_window": 128000,
    "supports_streaming": true,
    "supports_function_calling": true,
    "capabilities": {
      "supports_images": true,
      "supports_audio": false,
      "supports_video": false,
      "supports_documents": true,
      "supports_attachments": true,
      "supported_mime_types": ["image/png", "image/jpeg", "application/pdf", "text/plain"],
      "max_file_size_mb": 50,
      "max_attachments": 5
    }
  },
  "metadata": {
    "tenant_id": "…",
    "application_id": "…",
    "agent_type": "conversational",
    "created_at": "2026-05-02T18:11:04",
    "updated_at": "2026-07-01T09:30:12",
    "version": "v3",
    "api_compatibility": {
      "streaming": true,
      "attachments": true,
      "langchain_format": true,
      "function_calling": true
    }
  }
}

Notes:

  • configuration.environment_variables lists variable names only — values are never exposed. Use it to know which environment_overrides your streaming calls can supply.
  • configuration.tools are the tool identifiers the agent has attached.
  • agent_id is a string; treat it as opaque.

GET /v3/agent/models — what models are available

Lists every model your agent can run (for the model_name override on /v3/agent/stream), each with the same capability block as above:

{
  "models": [
    {
      "name": "claude-sonnet-5",
      "display_name": "Claude Sonnet 5",
      "provider": "anthropic",
      "context_window": 200000,
      "supports_streaming": true,
      "supports_function_calling": true,
      "capabilities": { "supports_images": true, "supported_mime_types": ["…"] }
    }
  ],
  "total_count": 18,
  "providers": ["anthropic", "openai", "google"],
  "default_model": "gpt-4o",
  "metadata": { "agent_id": "1042", "version": "v3", "supports_model_switching": true }
}

Models are sorted by provider, then name. default_model is the platform default used when an agent has no explicit model configured.

GET /v3/agent/capabilities — just the capability block

A convenience shortcut returning only the capabilities object from agent info — the fastest way to decide, client-side, whether to show an image-upload button or reject a file before uploading it:

{
  "supports_images": true,
  "supports_audio": false,
  "supports_video": false,
  "supports_documents": true,
  "supports_attachments": true,
  "supported_mime_types": ["image/png", "image/jpeg", "application/pdf", "text/plain"],
  "max_file_size_mb": 50,
  "max_attachments": 5
}

Check a file's MIME type against supported_mime_types before uploading; attachments the agent's model can't read are rejected when you try to use them in a streaming call.

Availability

All plans. Requires a deployed agent and an agent token.

Related articles