Skip to main content
This guide walks you through the three core steps to get up and running with the Agentic Framework API, starting with Authentication, then opening a thread with configured Experts and sending messages to that thread. We adhere to the A2A (Agent-to-Agent) specification, ensuring interoperability between Corti Agents and other systems built on the same open standard. In short, the A2A spec defines how agents can communicate, exchange context, and delegate tasks safely and consistently across different implementations.
Before you start…
  1. Get your API key from the Corti Console
  2. Configure Experts
  3. Integrate via REST or SDK in your product
  4. Follow guardrails – Human oversight is recommended for clinical use
1

🔐 Getting Credentials

  • Create a new OAuth client in the console.
  • Use the Client Credentials flow to obtain an access token.
  • Include the token in all requests using the Authorization: Bearer <token> header.
The Corti API Console is where you can create an account to access Corti AI   Create an account
   Create a project
   Create API clients
   Use the client ID and client secret in your app
   Develop against the API and collaborate with your project team
   Monitor usage, manage billing and more
2

🚀 Create an Agent

Use POST /agent to create a new agent. You can mix registry and remote Experts, each with their own configuration.Example:
POST /agents

{
    "name": "Example Agent",
    "systemPrompt": "Remeber you can explore more context in the data parts. Please send data part to the expert",
    "description": "Searches and retrieves structured information about drugs from reliable sources.",
    "experts": [
        {
            "type": "new",
            "name": "My_Custom_Expert",
            "description": "Use this for...",
            "systemPrompt": "You are a helpful expert specialized in...",
            "mcpServers": [
                {
                    "name": "testMCP",
                    "transportType": "streamable_http",
                    "authorizationType": "none",
                    "url": "https://example.com/mcp"
                }
            ]
        }
    ]
}
3

💬 Send a Message

Once you have an agent_id, send user input with POST /agent/{agent_id}/v1/message:send. The system routes the message to the right Experts and orchestrates the workflow.Example:
POST /agents/{{agent_id}}/v1/message:send

{
    "message": {
        "messageId": "your_unique_id",  // UUID
        "role": "user",
        "kind": "message",
        "parts": [
            {
                "kind": "text",
                "text": "Tell me about the latest guidelines for hypertension management."
            }
        ]
    }
}
4

... Next Steps

  • Passing context – Include conversation history for better responses
  • Receiving artifacts – Get structured data alongside text replies
Please contact us if you need more information about the Corti Agentic Framework.