Skip to main content
MCP Servers let you plug any Model Context Protocol server into your Coding Agent. Whatever tools that server exposes, the agent can call — as if they were built into pre.dev. If you’ve already built an MCP server for Claude, Cursor, or VS Code, it works here with zero changes.

​
When to use one

  • Internal tooling — your team has an MCP server that queries your staging database, triggers deploys, or exposes a private API
  • Third-party MCP servers — GitHub, Linear, Sentry, Grafana, Notion — any MCP-speaking service
  • Custom automations — any script you’ve wrapped as an MCP server (cron control, DNS changes, feature flag flips)

​
Connecting a server

  1. Open Integrations → MCP Servers
  2. Click Add MCP Server
  3. Paste the server’s JSON config — a remote server (url + headers) or a stdio server (command, args, env)
  4. Click Test — pre.dev handshakes with the server and lists its tools
  5. Save — the agent can now call those tools on any project

​
What the agent sees

Once a server is connected, its tools show up in the agent’s available toolset automatically. If your server exposes a queryStaging tool, the agent can reason about when to call it (“this task needs the latest customer count from staging → call queryStaging”) and execute the call during a build.

​
Toggling servers

MCP servers can be toggled on and off from the Integrations page (and with /mcp in the workspace or CLI). Turn off servers whose tools would add noise; leave on the ones that earn their place.

​
Security

  • Connection configs stay server-side and are never embedded in project code
  • Your server’s auth is respected — pre.dev forwards the headers you configured on every call