Q: Do I need the Buddy CLI?

For the skill-based setup, yes - install it with npm install -g bdy , run bdy login , then bdy proj link inside your project. The CLI reference documents every command the agent can reach. Pure MCP setups work without it.

Question 1

How does an AI agent connect to Buddy?

Accepted Answer

Two ways, and they work well together. The Buddy skill teaches the agent how to drive the bdy CLI - linking projects, deploying, running pipelines, reading logs. The remote MCP server gives it direct API access over the Model Context Protocol. Most setups use both.

Question 2

What is the Buddy MCP server?

Accepted Answer

An implementation of the Model Context Protocol hosted by Buddy, so an agent can actually do things in your workspace instead of just talking about them - trigger and monitor pipelines, manage deployments, spin up sandboxes, publish packages, route domains, and pull build output and logs. See the MCP docs for the server URLs.

Question 3

Which MCP server URL should I use?

Accepted Answer

Point your agent at the endpoint closest to your workspace region: https://mcp.buddy.works/mcp for US, https://mcp.eu.buddy.works/mcp for Europe, and https://mcp.asia.buddy.works/mcp for Asia.

Question 4

How does authentication work?

Accepted Answer

Interactive OAuth is the smoothest option for desktop clients - the agent opens a browser, you log in and approve the scopes. For CI pipelines and headless agents, use a personal access token passed as a Bearer header.

Question 5

Can I limit what the agent is allowed to do?

Accepted Answer

Yes. The server exposes over 90 tools across every category, and the ?tools= query parameter narrows them to a comma-separated list - for example ?tools=pipelines,source. Start with a narrowly-scoped token and read-only access, then expand as you build trust. Details in the MCP docs.

Question 6

Do I need the Buddy CLI?

Accepted Answer

For the skill-based setup, yes - install it with npm install -g bdy, run bdy login, then bdy proj link inside your project. The CLI reference documents every command the agent can reach. Pure MCP setups work without it.

Question 7

Can I run an agent inside Buddy instead of on my machine?

Accepted Answer

Yes. Agents can run in a fully isolated Buddy sandbox on a real, shareable URL, or as an action inside a pipeline. You can also self-host Claude Managed Agents on Buddy sandboxes.

Question 8

My agent is not on the list - will it still work?

Accepted Answer

Most likely. Any MCP-compatible client can connect to Buddy with a standard mcpServers entry pointing at the server URL, and the Buddy skill works with any agent that can run shell commands. The agents docs cover the tool-specific config files.

AI coding agents

Claude Code

Antigravity

Codex

Cursor

OpenCode

GitHub Copilot

Pi

Devin

Kiro

Grok Build

FAQS