Zed MCP Server Setup: Connect Tools to Your Coding Assistant

Zed is a code editor with an integrated AI assistant. It supports MCP configuration, which means you can add outside servers that give the assistant access to tools, data sources, and services beyond the built-in ones. Setting up an MCP server in Zed involves editing a JSON configuration file and verifying that the tools appear and work as expected. Tools like fenbs let you carry the project decisions and progress that survive when you move between Zed and other AI coding tools.

MCP, the Model Context Protocol, is an open standard for connecting AI assistants to outside tools. Zed uses the same configuration approach as other clients, so a server that works in Claude Desktop or Cursor can usually be added to Zed with the same settings.

Where Zed stores MCP configuration

Zed reads MCP server configuration from its settings file. On most systems, the settings file is at ~/.config/zed/settings.json. Open it in Zed or in another editor, then add or update the context_servers section with your server entries. Each entry needs a name, the command to run the server, and any arguments or environment variables it requires.

"context_servers": {
  "fenbs": {
    "command": "npx",
    "args": ["-y", "fenbs-mcp"]
  }
}

Save the file and restart Zed or reload the settings. Open the assistant panel and check whether the tools from the new server appear in the tool list. Zed shows the connected tools in the assistant context; if a tool does not appear, check the server log for an authentication or startup error.

Checking tool availability and permissions

Before using a tool in a coding session, confirm it behaves as expected on a small test request. Ask the assistant to call the tool with a known input and verify the result. This catches authentication errors, scope limits, or API version mismatches before they affect real work.

For servers that write to an outside system, review what the tool can change and whether it requires approval. Zed's assistant can use tools during a conversation; a write tool that acts without confirmation can make changes that are hard to reverse. Start with read-only access and add write tools only when you have confirmed the output of the read tools is accurate.

Using MCP servers during a coding session

With MCP servers configured, the Zed assistant can call tools when they are relevant to a question. Ask it to read a project context before starting a feature, or to check a database schema before writing a query. The assistant uses the tool result as context for its answer without you having to paste the information manually.

Give the assistant a clear starting instruction when you begin a session on an unfamiliar part of the codebase. Ask it to read the relevant project notes and decisions before proposing changes. That starting step reduces the chance of an assistant working from outdated assumptions.

Carrying project context when you switch AI coding tools

Many developers use more than one AI coding tool. You might use Zed for focused editing and switch to Claude Code or Cursor for longer tasks. Each tool starts a fresh session, and the project decisions made in one session are not automatically available in another.

A shared project record outside any one tool solves this. The decisions, current tasks, and open questions live in one place that both Zed and other supported tools can read over MCP. When you switch tools, the next assistant reads the same context and continues from the same point rather than asking you to repeat the background.

See mcp-server-configuration for general configuration patterns that apply across clients, and mcp-server-hosting for options when you need a shared server that multiple team members can use.

Troubleshooting a Zed MCP connection

If tools do not appear after saving the settings file, check the context_servers key spelling and JSON syntax first. A single missing comma prevents the file from parsing. Then open the Zed log or developer console for startup errors from the server process.

For authentication errors, confirm the server's required environment variables are present and that the token or API key is valid. Some servers expect the credentials as environment variables in the settings entry rather than in the shell environment. Check the server's README for the exact format.

How to do this with fenbs

1. Create a free account and add a project for the codebase you are working on.
2. Add the fenbs MCP server to Zed's context_servers configuration in ~/.config/zed/settings.json. Follow the setup guide at Integrations for the exact entry.
3. At the start of each session, ask the Zed assistant to read the project context: Rules, Where we left off, and the current tasks. Confirm it can see the right project before you start coding.
4. Record architectural decisions and coding conventions in Decisions and rules so every session in Zed and in other tools starts from the same approved baseline.
5. Before closing a session, ask the assistant to save Where we left off with the tasks it touched and the next step. Open the next session in any connected tool and ask it to read that note before starting.