Agent SDK overview

The Corti Agent SDK is in alpha v2 private preview. The API surface may change between releases, and the Python package is still being rebuilt for v2 parity. Do not use it in production without a dedicated support arrangement. Contact help@corti.ai to request access.

The Corti Agent SDK (@corti/agent-sdk for TypeScript, corti-agent-sdk for Python) is a developer-friendly wrapper for building multi-agent systems on the Agentic Framework v2 API. The TypeScript package is the v2 reference implementation. The Python package is being rebuilt for v2 parity; until then, the two may differ in available features. The TypeScript package is built on openapi-ts with openapi-fetch as its only runtime dependency. The Python package uses httpx with async/await.

What it provides

Agent lifecycle: create, list, get, update, and delete agents via typed resource clients
A2A messaging: send messages, stream responses, and manage conversation contexts with automatic context ID tracking
Connector factories: build registry, mcp, agent, a2a, and schema (TypeScript only) connectors with typed factory helpers
Multi-agent composition: deterministic workflow pipelines, parallel fan-out, and stateGraph routing with cycles and typed shared state
Structured errors: typed error classes (CortiError, ManagementError, HttpError, A2AError) for both management and A2A planes
Extended resources: contexts, registry, usage, feedback, agent cards, and models

Prerequisites

A Corti API tenant with credentials
TypeScript: Node.js 18 or later (uses native fetch and AsyncIterable)
Python: Python 3.9 or later

The TypeScript package is ESM-only. For CommonJS projects, use dynamic import() or set your project to ESM.

Install

npm install @corti/agent-sdk

pip install corti-agent-sdk

Quick example

import { CortiClient, AgentHandle, connectors } from "@corti/agent-sdk";

// Replace these with your values
const REGISTRY_NAME = "@corti/medical-coding";
const TENANT = "<your-tenant-name>";
const TOKEN = "<your-access-token>";

const client = new CortiClient({ token: TOKEN, tenant: TENANT });

const agent = await client.agents.create({
  name: "my-agent",
  description: "Handles medical coding queries",
  connectors: [connectors.registry(REGISTRY_NAME)],
});

const handle = new AgentHandle(agent, client);
const ctx = handle.createContext();
const reply = await ctx.sendText("ICD-10 code for hypertension?");

console.log(reply.text); // "The ICD-10 code is I10."

import asyncio
from corti_agent_sdk import CortiClient, AgentsClient, connectors

# Replace these with your values
CLIENT_ID = "<your-client-id>"
CLIENT_SECRET = "<your-client-secret>"
ENVIRONMENT = "<eu-or-us>"
REGISTRY_NAME = "@corti/medical-coding"
TENANT = "<your-tenant-name>"


async def main():
    async with CortiClient(
        tenant_name=TENANT,
        environment=ENVIRONMENT,
        auth={"client_id": CLIENT_ID, "client_secret": CLIENT_SECRET},
    ) as client:
        agents = AgentsClient(client)

        agent = await agents.create(
            name="my-agent",
            description="Handles medical coding queries",
            connectors=[connectors.registry(name=REGISTRY_NAME)],
        )

        ctx = agent.create_context()
        reply = await ctx.send_text("ICD-10 code for hypertension?")
        print(reply.text)  # "The ICD-10 code is I10."


asyncio.run(main())

Decision guide

You want to…	Use
Send one message and get one reply	`agent.run(text)` or `ctx.sendText`
Hold a multi-turn conversation	`agent.createContext()` + `sendText`
Resume a thread from a previous process or session	`agent.getContext(contextId)`
Stream tokens as the agent writes them	`ctx.streamMessage([...])`
Run a fixed pipeline of agents (A to B to C)	`workflow([...])`
Run agents concurrently on the same input	`parallel([...])`
Conditionally branch, loop, or share typed state across agents	`stateGraph<State>()`
Attach an MCP server, sub-agent, or registry connector	`connectors.*` in `create({...})`

Rule of thumb: start with workflow. Only reach for stateGraph when you need cycles or branching that depends on accumulated state.

Next steps

Core concepts

Learn about the client, agents, contexts, connectors, message responses, and streaming.

Multi-agent composition

Build workflows, fan out in parallel, and route with state graphs.

API reference

Full method and type reference for both languages.

Common pitfalls

Avoid the most frequent mistakes when using the SDK.

Quickstart

Concepts

Guides

Reference

Registry Connectors

Agent SDK

v1 Documentation (archived)

What it provides

Prerequisites

Install

Quick example

Decision guide

Next steps

Core concepts

Multi-agent composition

API reference

Common pitfalls

​What it provides

​Prerequisites

​Install

​Quick example

​Decision guide

​Next steps

Core concepts

Multi-agent composition

API reference

Common pitfalls

What it provides

Prerequisites

Install

Quick example

Decision guide

Next steps