Skills & MCP

Two kinds of add-ons make an agent better at a job, and NeuroSquad installs both for you:

How it works

Which agents can use them

The card says the same thing next to each connected agent: Live, Next start or Can’t use. The last one means NeuroSquad doesn’t give tools to that agent at all — the same agents that arrows don’t give abilities to.

How the agent knows what it has

You don’t have to tell the agent anything.

• Skills: the agent’s skill_load tool lists every connected skill with its “when to use it” line, so the agent loads the matching skill on its own before it starts a task — the same way Claude Code’s own skills work.
• MCP servers: each server’s tools appear under the server’s name, like framelink__get_figma_data, with the server named in every description, plus the server’s own usage notes.
• neurosquad_connections — the tool an agent calls to see what is on the other end of its arrows — lists the connected servers and skills too.

Agents equip themselves

An agent doesn’t have to wait for you to connect things. It knows what NeuroSquad offers and can add it to its own card when the task needs it:

• It knows the list. On the first turn of a session the agent gets a short note — the plugins that work for it and the skills and MCP servers you installed, one line each (about 300 tokens at most). The tool neurosquad_catalog gives the details: what each one is for, the tools it brings, network and setup notes, and whether it is already connected.
• It connects what it needs. neurosquad_equip puts the card beside the agent (or reuses the one already on the canvas) and draws the arrow — the tools are there right after. You see a notice, for example “Claude Code connected Context7”, with Undo.
• New installs always ask you. A skill or MCP server that is not installed yet is installed only after you press Install and connect in a dialog that shows exactly what would be installed: the package and version, the source, the commands, the pinned hash. Nothing installs automatically — not in dangerous mode either. A server that needs a key or an address opens in the catalog for you to set up. Plugins that download something (RTK, Code Graph, Graphify) still wait for you to start the download on their card.
• Dangerous mode: you allow MCP servers. An agent in dangerous mode would run an MCP server’s tools without asking you each time, so it can’t connect one by itself. Instead you get a dialog: which agent, which server, the tools it would get and how the server starts — Allow connects it (as if you had drawn the arrow), Deny is remembered until the app restarts. If the server isn’t installed yet either, it is one dialog: Install and allow.
• The usual rules still apply. Each call to an MCP server’s tools is approved as always; an agent can’t take over a terminal, browser or MCP card you set up; RTK doesn’t work in WSL/SSH workspaces, Graphify not over SSH.
• CLIs that list their tools once. Codex, Kimi Code, ZCode, PenguinHarness, Cursor and Crush read their tool list at start: plugin and skill tools are listed for them from the start and work as soon as the arrow exists, but a newly connected MCP server’s tools reach them only after the card restarts (the conversation resumes).

To turn it off, switch off Agents can add plugins, skills and MCP to themselves in the workspace’s edit dialog. Then only you connect them.

Things to know

• You approve each call to an installed server. Its tools come from a separate server entry (ns-connected) that NeuroSquad does not pre-approve, so Claude Code, Codex and Qwen ask before each call — unless you switched your agent to a mode that approves everything itself. Hermes Agent never asks on its own, so for it NeuroSquad asks: Allow once, Allow for this session or Deny (no answer in five minutes counts as Deny). A server that later changes its tools shows “This server changed its tools” on its card; new and changed tools stay hidden from agents until you press Accept changes.
• It’s third-party code. A server you install runs on your computer (or is a service on the internet); a skill’s text goes straight into the agent’s instructions. Install what you trust — the catalog shows the source, the exact commands and, for skills, security scans.
• Not every server can be installed yet. Servers published only as MCPB bundles, NuGet or Cargo packages, and packages that start a local web server, are shown but not installable.
• Some sign-ins only allow approved apps. Figma’s official remote server, for example, refuses sign-in from apps it hasn’t approved. Use a server that takes a personal token instead — Framelink Figma MCP (FIGMA_API_KEY) works with your own Figma token.