Bring your own tools, local or remote.

MCP servers extend the tools available to the model beyond what ships in the box. A connected server’s tools, its own instructions, its resources and its resource templates are all made available to the agent.

Install ML.ai CodeBook a call

MCP servers

Connect local or remote tools that ML.ai Code can use while working.

No MCP servers configured.

Add server

How it works

Connect tools over the Model Context Protocol, scoped per project or globally, with environment variables, headers, timeouts and OAuth sign-in.

01

Local servers

Command-based, for example npx -y package-name, with environment variables one KEY=value per line and a configurable startup and request timeout.

02

Remote servers

An HTTP or HTTPS URL, with request headers, a timeout, and OAuth sign-in including optional client ID, client secret and scope.

03

Project or global scope

Project scope writes to the repository’s configuration and travels with it. Global scope is for personal tools you want in every project.

04

Full lifecycle

Connect, disconnect, sign in, sign out, and permanently remove. /mcps reports how many configured servers need attention.

Adding a local server, scoped to one project

    1

    You open the form

    /mcps

    2

    You fill in type, scope, and command

    Type: Local. Scope: Project. Command: npx -y @some-org/mcp-server-example.

    3

    Project scope decides who gets it

    Scoping it to Project rather than Global means this server is written to the project’s own configuration and only starts for people working in this repo, not in every project you open.

    4

    npx -y skips the install step

    It runs the package without a separate install, fetching it fresh each time the command starts. /mcps reports how many configured servers need attention if anything fails to connect.

MCP servers

Connect local or remote tools that ML.ai Code can use while working.

TypeLocal
ScopeProject

Command

npx -y @some-org/mcp-server-example

Worth knowing

  • A server reporting a successful connection is not proof its tools work. If a server matters, invoke one of its tools and check the result.

  • Do not commit secrets to project configuration. Prefer environment-variable references or OAuth.

  • MCP tool input is sent to the MCP server chosen for that operation.

Questions worth asking

Related

Skills

Type / to attach a skill to your next message. Five ship with the product, and your own project and global skills appear in the same list.

Read more

Agents

Choose a subagent for the composer and it stays selected until you remove or replace it. Each one is built for a different kind of work, and only general carries out changes end to end.

Read more

Try ML.ai Code today, or talk to us about what is next.

Install the editor agent on your own machine, or book a call to talk through your team's workloads.

InstallBook a call