Executor

Executor

Connect any agent to everything.

Executor is an open-source integration layer for AI agents. Configure everyintegration once (MCP servers, OpenAPI specs, GraphQL APIs) with authenticationand per-tool policies, then use that one catalog from any MCP-compatible agent.

Website · Documentation · Discord

https://github.com/user-attachments/assets/11225f83-e848-42ba-99b2-a993bcc88dad

Why Executor

Every agent you use (Claude Code, Cursor, ChatGPT, and the rest) needs its owncopy of every integration: the same API keys pasted in three places, the sameMCP servers wired up again, no shared idea of what each tool is allowed to do.

Executor is the layer in between. Add a tool once, give it credentials once,set its policy once, and every agent shares it over MCP. Your integrations,auth, and policies live in one place instead of being scattered across eachclient.

• Any source. First-party support for MCP servers, OpenAPI, GraphQL, andGoogle Discovery. If you can describe it with a JSON schema, it can be anintegration. The plugin system is open to any source type.
• One catalog, every agent. Anything MCP-compatible connects to the sameset of tools.
• Governed by policy. Each tool is allowed, gated behind approval, orblocked, with sensible defaults derived from the spec.
• Run it your way. Local CLI, a desktop app, hosted Executor Cloud, orself-hosted on Docker or Cloudflare. Same functionality, different packaging.

How it works

1. Add an integration: an MCP server, an OpenAPI spec, or a GraphQL API.
2. Create a connection: one configured (optionally authenticated) instanceof that integration. An integration can have many connections.
3. Set policies: decide whether each tool is always allowed, needsapproval, or is blocked.
4. Point your agents at Executor over MCP. They all share the same catalog.

See Concepts for the full model.

Quick start

The fastest path is Executor Cloud: sign in, add anintegration, and point your agents at the hosted MCP endpoint. Nothing toinstall.

To run it locally instead (Node.js 20+):

npm install -g executor   # or: pnpm add -g / bun add -g / yarn global add
executor install          # install the durable background service
executor web              # open the web UI in your browser

executor install keeps the service running across restarts. For a throwawayforeground runtime, use executor web --foreground. From the web UI, add yourfirst integration and connect an agent.

Set up with your agent

Prefer to let your coding agent do the setup? Copy thesetup prompt from the docs and paste it into Claude,Cursor, or any MCP-capable agent. It will help you pick how to run Executor,install it, connect over MCP, and get your first integration working end to end.

Ways to run

Every form exposes the same functionality, just packaged differently.

Connect an agent over MCP

Add Executor to any MCP client (Claude Code, Cursor, OpenCode) withadd-mcp, which detects the client andwrites its config for you:

# Over HTTP (the running service serves a streamable-HTTP endpoint)
npx add-mcp http://127.0.0.1:4788/mcp --transport http --name executor

# Or over stdio, with the executor CLI on your PATH
npx add-mcp "executor mcp" --name executor

The Connect card in the web UI shows the exact command (and port, if itdiffers) already filled in. Most MCP clients only load servers at startup, soyou may need to restart the client or open a new chat before the Executor toolsappear.

Add an integration

From the web UI, click Add Source, paste an OpenAPI, GraphQL, or MCP URL,and Executor detects the type, indexes the tools, and handles auth. Or add onefrom the CLI:

executor call executor openapi addSource '{
  "spec": "https://petstore3.swagger.io/api/v3/openapi.json",
  "namespace": "petstore",
  "baseUrl": "https://petstore3.swagger.io/api/v3"
}'

Use baseUrl when the OpenAPI document has relative servers entries (forexample "/api/v3"). Confirm it is live with executor tools sources.

Using tools

From the CLI

executor tools search "send email"      # find tools by intent
executor call github issues --help      # browse a namespace
executor call github issues create '{"owner":"octocat","repo":"Hello-World","title":"Hi"}'

executor call, executor resume, and executor tools ... auto-start thelocal daemon if needed, and pick a free port if the default is busy. If anexecution pauses for auth or approval, resume it:

executor resume --execution-id exec_123

From your own code

Embed Executor with the TypeScript SDK (a Promise API; an Effect-native API isalso available):

import { createExecutor } from "@executor-js/sdk/promise";
import { openApiPlugin } from "@executor-js/plugin-openapi/promise";

const executor = await createExecutor({ plugins: [openApiPlugin()] });

// add an integration, create a connection, then list and call tools
const tools = await executor.tools.list({ integration: "inventory" });
const schema = await executor.tools.schema(tools[0].address);

await executor.close();

See examples/ for runnable end-to-end scripts.

CLI reference

executor install                    # install/start the durable background service
executor web                        # open the running web UI
executor web --foreground           # start a temporary foreground runtime + web UI
executor daemon run                 # start persistent local daemon in background
executor daemon status              # show daemon status
executor daemon stop                # stop daemon
executor daemon restart             # restart daemon
executor mcp                        # start MCP endpoint (stdio)
executor call <path...> '{"k":"v"}' # invoke a tool by path segments
executor call <path...> --help      # browse namespaces/resources/methods
executor call <path...> --help --match "<text>" --limit <n> # narrow huge namespaces
executor resume --execution-id <id> # resume paused execution
executor tools search "<query>"     # search tools by intent
executor tools sources              # list configured sources + tool counts
executor tools describe <path>      # show tool TypeScript/JSON schema

Project layout

Executor is a Bun + Turborepo monorepo.

apps/
  cli/             the `executor` CLI and local background service
  desktop/         the desktop app (Mac, Windows, Linux)
  local/           the local runtime shared by the CLI and desktop
  cloud/           Executor Cloud (the hosted product)
  host-selfhost/   self-hosted server (Docker)
  host-cloudflare/ Cloudflare Worker deployment
  marketing/       the executor.sh site
  docs/            the docs at executor.sh/docs
packages/
  core/            contracts, plugin wiring, scopes, policies, SDK, API, CLI core
  kernel/          execution runtimes (QuickJS, Deno subprocess, dynamic worker)
  plugins/         source and provider plugins (openapi, graphql, mcp, google,
                   microsoft, 1password, keychain, encrypted/file secrets, ...)
  hosts/           host adapters (MCP surface, Cloudflare)
  react/           shared React UI
  app/             the web app UI
examples/          runnable SDK examples
e2e/               full-stack end-to-end tests

Develop locally

bun install
bun run bootstrap   # idempotent: install deps, build required artifacts, fetch Playwright
bun run dev         # start the dev servers (defaults to http://127.0.0.1:4788)

bun run bootstrap is required in a fresh checkout: its build step producesartifacts the dev servers fail without. See RUNNING.md for devservers, ports, and environment gotchas, and AGENTS.md for thecontributor contract.

Tests

bun run test       # unit + integration suites
bun run test:e2e   # full-stack e2e: boots the cloud and self-host apps and drives them

The browser e2e scenarios need Playwright's Chromium once per machine:bunx playwright install chromium.

Community

Join the Discord. To learn more, visitexecutor.sh or Ask DeepWiki.

License

MIT

Attribution

• Thank you to Crystian for providingthe npm package name executor.

References

As part of my coding process, I give my agent access to references to othercodebases to understand patterns and how other people have implemented systems.A non-exhaustive list:

• FumaDB - Storage adapter reference
• Effect - General code patterns
• OpenCode - Plugin system reference
• OpenClaw - Plugin system reference
• Emdash - Plugin system reference
• Pi - Plugin system reference

You are also encouraged to use this codebase as a reference to understand how itis implemented.