Integrations

MCP servers

Connect Model Context Protocol servers to give your agents extra tools — a filesystem, a database, GitHub, or anything else that speaks MCP. Enabled servers are handed to Claude Code automatically.

Back to app

General

Models

Agents

Workflows

MCP servers

Commands & Skills

Appearance

Git

Account

About

v0.1.0-beta.1

MCP servers

Add server

Add local (stdio) or remote (SSE/HTTP) servers. Enabled servers are passed to every agent via --mcp-config.

filesystemstdio

npx @mcp/server-filesystem ~/code

githubremote

https://mcp.github.com/sse

postgresstdio

uvx mcp-server-postgres

# written on save, then handed to Claude Code

claude --mcp-config ~/.neuralcode/mcp.json

Settings → MCP servers: add local or remote servers, then toggle the ones you want on.

Adding a server#

Open Settings → MCP servers and click Add server. There are two kinds:

Kind	How it's defined	Use for
Local (stdio)	A `command` with `args` and optional `env` — NeuralCode launches the process and talks to it over stdio.	Tools that run on your machine — filesystem, a local database, custom scripts.
Remote (SSE / HTTP)	A `URL` the server is reachable at.	Hosted services — a team server, a SaaS integration.

1
Choose local or remote
Pick the type, then fill in the command/args/env or the URL.
2
Name it
The name is how the server shows up in the list and in tool calls.
3
Enable it
Flip the toggle on. You can enable, edit, or delete any server later.

On save, NeuralCode writes every enabled server to a config file and passes it to Claude Code with the --mcp-config flag. That means the tools are available to every agent in a workspace — including each stage of an agent pipeline — with no per-task setup.

claude --mcp-config ~/.neuralcode/mcp.json …

Disabled servers are simply left out of the config, so turning one off removes its tools from the next run.

Servers run with your trust

A local server runs as a process on your machine and a remote server receives whatever the agent sends it. Only add servers you trust, and scope filesystem servers to the directories you actually need.

Note

Permission still applies: an agent in read-only or plan-modecan call MCP tools to read, but won't be allowed to make changes through them.

GitHub issues → agent

Google Sign-In

MCP servers

MCP servers

Adding a server#

How it reaches the agent#