# Claude Cowork has no documented task API. An open-source alternative does > Cowork has no documented task API. The feature runs inside the Claude app, and the APIs Anthropic documents around it read usage and transcripts on Enterprise plans. Kortix is the open-source CompanyOS, and your code can call the system your team already works in, which is my pick when you need an API; I work on Kortix's marketing, so weigh that accordingly. URL: https://claudecoworkalternative.com/claude-cowork-api ## Four ways into the same session HTTP under /v1, an MCP server at api.kortix.com/v1/mcp, a typed TypeScript client, and the chat channels: Slack, Microsoft Teams and email. ### REST API `docs.md` ``` The Kortix REST API is the HTTP interface to accounts, projects, sessions, secrets, connectors, triggers, and Apps. The web app, the CLI, the SDK, and the Kortix MCP server all call it. ``` Source: https://kortix.com/docs/developers ### MCP server `docs.md` ``` The Kortix MCP server lets another AI tool, such as Claude, ChatGPT, Cursor or Codex, use Kortix. You install nothing: you add one URL to the tool and sign in. The tool acts as you, with your permissions, and reaches every account and project you can open. ``` Source: https://kortix.com/docs/guide ### SDK `docs.md` ``` @kortix/sdk is the TypeScript client for the Kortix API. It creates projects and sessions, sends prompts, streams the agent's reply and reads the result. ``` Source: https://kortix.com/docs/developers ### Slack, Teams and email `docs.md` ``` Mention the bot with a task The agent replies in the thread, with live progress. ``` Source: https://kortix.com/docs/guide ## Programmatic access, side by side Five rows on how a task starts, with the Cowork answers taken from Anthropic's pages, checked October 2026. | | Kortix (recommended) | Claude Cowork | | --- | --- | --- | | Documented way to start a task | REST under /v1; reference at api.kortix.com/v1/docs | None documented; tasks start in the Claude app | | Other programmatic surfaces | MCP server at /v1/mcp; typed SDK; CLI | Connectors, skills and plugins it calls, no task endpoint | | Separate build-it-yourself product | Not needed; call the live system | Managed Agents, a separate product billed per token and session-hour | | Usage and compliance APIs | Audit log keeps 90 days of events live | Analytics and Compliance APIs, Enterprise plans only | | Open source | Open source. Read it, fork it, audit it | No source published | ## Three things the word API names around Cowork [Anthropic's product page](https://claude.com/product/cowork) puts Cowork inside the Claude app, on web, desktop and mobile (checked October 2026). A task starts from the message box beside Chat, and the finished work comes back for review in Claude. The [Cowork docs](https://claude.com/docs/cowork/overview) list what extends it: connectors over MCP, skills, and plugins, each called by the assistant while it works. Neither page describes an endpoint you could call to hand Cowork a task. The second sense lives one product over, on the Claude Platform. [Managed Agents](https://platform.claude.com/docs/en/managed-agents/quickstart) is Anthropic's build-it-yourself agent product: you define the agent, the sandbox environment and the session, and your code drives the loop. The Agents, Sessions and Environments endpoints are in beta, per [the API overview](https://platform.claude.com/docs/en/api/overview). Billing has two lines: tokens at standard model rates, and $0.08 per session-hour of running time, metered to the millisecond. Anthropic's own worked example prices a one-hour coding session with 50,000 input and 15,000 output tokens at $0.58 all in ([the pricing docs](https://platform.claude.com/docs/en/about-claude/pricing), checked October 2026). The third sense is reporting. The Enterprise row of the Cowork product page promises admin controls, usage analytics, the Analytics API and OpenTelemetry observability. The [pricing table](https://claude.com/pricing) puts the Compliance API on Enterprise plans alone, and its documentation says the session endpoints return transcripts of Cowork sessions. The [Analytics API](https://platform.claude.com/docs/en/manage-claude/analytics-api) reports engagement, adoption and cost data across Claude products for Enterprise organizations. All three read records of work already done. None of them starts a task. What a developer usually means by an API is the one thing none of the three provides: start this task, stream the progress, get the result back, from my own code. The platform product builds a new agent rather than reaching the one your team already uses. The Enterprise APIs watch the app without driving it. ## What each one is for The app feature suits a person handing off a task from their own desk. It schedules recurring work, runs in Anthropic's cloud or in your Amazon Bedrock, Gemini Enterprise Agent Platform or Microsoft Foundry account, and returns the result inside Claude. A team that wants its own systems to trigger the assistant finds no documented way in. Managed Agents suits a developer building a product of their own. You assemble the agent, pick the model, wire the tools, and your application owns the interface. Nothing carries over from the Cowork your team works in, and the session-hour line bills on top of tokens. Anthropic's [release notes](https://support.claude.com/en/articles/12138966-release-notes) point the same way: the October 2026 note that gives Max and Team plans monthly API credits describes them as for running your own apps and agents on the Claude Platform. The observability APIs suit an IT or security team that has to answer for usage. They pull reports and transcripts for review and compliance work. If the plan was to make Cowork a callable part of your own stack, that layer is where Anthropic's documented surface stops. ## What calling the system you own looks like Kortix stands on the other side of that gap. It is open source, and the system your team works in is the system your code can call. A REST API under /v1 starts sessions, with its interactive reference at api.kortix.com/v1/docs ([Read the docs](https://kortix.com/docs)). An MCP server at /v1/mcp serves any MCP client: it starts sessions, reads files and calls connectors as you. For TypeScript, the typed @kortix/sdk client wraps the API and each session's runtime. Every surface starts the same session, so an API-triggered run looks to the system like a colleague's desk task. There is no second product to buy and no separate bill: the surface your code calls is the one your team already works in. Ownership reaches the details a platform lead asks about. Connector credentials are brokered server-side and never enter the machine, and the path from your API call into a session is code you can read and change, because the whole system is open source. Models stay your choice. Any model provider with your own keys. Or the ChatGPT plan you already pay for. Set the model on an agent, on one session, or on a single message, and switch when a better one lands. ## Questions before you integrate Dated answers; prices and availability checked October 2026. ### Does Claude Cowork have an API? No. Anthropic's product and help pages document no task API for the app: tasks start in the Claude app, and the APIs around it read usage and transcripts on Enterprise plans. The build-it-yourself path is Managed Agents on the Claude Platform, a separate product. ### What is the difference between the Cowork Analytics API and an agent API? The Analytics API reads. It reports engagement, adoption and cost data for Enterprise organizations. An agent API acts: it starts a session and receives the finished work. Kortix's REST API under /v1 starts sessions; the Analytics API only reports on them. ### Can I call Kortix agents from my own app? Yes. A REST API under /v1 starts and follows sessions, an MCP server at /v1/mcp serves any MCP client, and @kortix/sdk is the typed TypeScript client. Every surface starts the same session, so an API call reaches the same system your team works in. ### What does a Managed Agents session cost? Tokens at standard model rates plus $0.08 per session-hour of running time, metered to the millisecond. Anthropic's worked example: a one-hour coding session with 50,000 input and 15,000 output tokens costs $0.58 (checked October 2026). ## Wire your code to the system your team works in A REST call, an MCP client and a Slack mention all reach the same session. [Try Kortix](https://kortix.com) · [Read the docs](https://kortix.com/docs) --- A guide written by someone who works on Kortix's marketing. My pick is Kortix: open source, one repo, free to self-host. Revised 2026-11-06 Not affiliated with Anthropic, OpenAI or Microsoft. Claude Cowork, ChatGPT Work and Copilot belong to their owners. © 2026 claudecoworkalternative.com