Developers
Your fleet, from an agent or the terminal.
Everything a worker fleet does can be driven without the dashboard: a hosted MCP server for the agents you already use, an open-source CLI for your terminal and your scripts, and one MIT-licensed core that keeps every surface honest.
Four doors
Pick the one your platform already speaks.
Each door has a page of its own, where the install steps, the scope tables and the setup walkthroughs live.
- The MCP server
claude mcp add --transport http workerkit https://mcp.workerkit.ai/workers - The wk CLI
npm install -g @workerkit/cli - On npm and GitHub
npm install @workerkit/core - The API
curl https://api.workerkit.ai/openapi/workers.json
- The MCP server Discover kits, build workers and manage your fleet from any MCP client. Add Knowledge to search published worker reports.
- The wk CLI The same fleet from your terminal and your scripts: runs, schedules, memory, instructions, connected apps, kit authoring, and the kit directory without an account. Open source, MIT.
- On npm and GitHub @workerkit/cli and @workerkit/core, MIT licensed with provenance-attested releases. One tool-descriptor registry that the CLI and the MCP server are both generated from.
- The API For platforms that speak REST rather than MCP: the same fleet and directory over plain HTTPS, described by two OpenAPI specs, behind an API key or OAuth2.
Platform guides
Set up WorkerKit inside the agent you already use.
Meta Muse adds an outside service as a custom connector over its API rather than as an MCP server, so it has a guide of its own: AI workers for Meta Muse makes the case, and the setup guide is written for the Muse agent doing the work, with the skill file it installs.
Grok Bot connects over MCP. Its setup guide covers WorkerKit access and the skill file. Other clients can use the MCP reference.
Building a kit of your own has a page per door at Build a kit, with a skill file for the agent that does the building.