8 Command Code MCP Servers Worth Trying

4 min read

,
Best MCP servers for Command Code

If you use Command Code, you do not need to build every integration yourself. Its official MCP guide lists servers for documentation, GitHub, browsers, errors, design and team notes.

The server details and setup commands come from Command Code’s official documentation.

Command Code is a terminal (and desktop) coding agent built for open models. MCP (Model Context Protocol) is how it reaches tools outside the repo: Notion, Sentry, Figma, live docs, and more. You add a server with cmd mcp add, authenticate if asked, then ask in plain language.

Command Code coding agent interface for MCP tools and plugins
Command Code can call MCP tools beside its built-in file and shell tools. Image: Command Code.

MCP servers are not Skills or mods

Command Code splits extensions into a few layers:

  • MCP servers: connect Notion, GitHub, browsers, and other services
  • Skills: reusable instruction packs (cmd skills add)
  • Mods: TypeScript packages that change agent behavior (cmd mods add)
  • Custom commands: your own slash commands under .commandcode/commands/

Best MCP servers for Command Code

Command Code documents these eight servers and their setup commands. Start with the servers you will actually use. A long list adds clutter and gives the agent access to more tools than the task needs.

1. Context7 (fresh library docs)

Context7 supplies version-specific library documentation and examples, which helps the agent work from current APIs.

cmd mcp add context7 -- npx -y @upstash/context7-mcp

Tip from the docs: include use context7 in prompts when you want latest library docs.

2. OpenAI Docs

Remote HTTP server for OpenAI developer documentation inside the session.

cmd mcp add --transport http openai-docs https://developers.openai.com/mcp

3. Notion

Search, read, and create pages in your Notion workspace. Good for specs and PRDs the agent should not invent.

cmd mcp add --transport http notion https://mcp.notion.com/mcp

Authenticate via /mcp after adding.

4. GitHub

GitHub MCP can work with issues, pull requests and code search. It requires a personal access token; use a fine-grained token with only the permissions you need, and do not commit a real token to a shared config.

Docs walk through fine-grained tokens and a sample cmd mcp add-json github ... payload pointing at GitHub’s hosted MCP endpoint.

5. Sentry

Bring production errors and traces into the agent so fixes start from real stack traces.

cmd mcp add --transport http sentry https://mcp.sentry.dev/mcp

Authenticate in /mcp.

6. Playwright

Drive a real browser for UI checks, screenshots, and web automation from the CLI agent.

cmd mcp add playwright -- npx -y @playwright/mcp@latest

7. Chrome DevTools

Inspect and control Chrome for DOM debugging. Chrome must run with remote debugging on port 9222.

cmd mcp add chrome-devtools -- npx -y chrome-devtools-mcp@latest

8. Figma

Design context for build-from-mockup work. Remote HTTP server, or a local server tied to the Figma desktop app.

cmd mcp add --transport http figma https://mcp.figma.com/mcp

Remote needs OAuth via /mcp.

Bonus from the same docs: Stripe

Command Code’s MCP examples also show Stripe’s remote server (https://mcp.stripe.com) when explaining scopes. Use it if payments live in your stack; skip it if they do not.

How to add and check an MCP server

Three scopes matter: local (default, per project), project (.mcp.json, shareable), and user (all projects). List with cmd mcp list, inspect with cmd mcp get, and open status inside a session with /mcp.

Already set up MCP in Claude Code, Cursor, OpenCode, or Gemini CLI? Run /import (or /import opencode, etc.) to pull servers, skills, and commands across.

What should you install first?

Start with Context7 and the one service your team actually uses as its source of truth, such as Notion or GitHub. Add Sentry when you work with production errors. Install Playwright or Chrome DevTools only if the agent needs to operate a browser. Keeping the list short makes it easier to manage permissions and tool access.

Use a mod when you want to add or change agent behaviour; use a Skill for reusable instructions. Command Code’s mods docs and Skills docs explain the setup.

Share this story

About the author