MCP Tools
SkillAI & modelsMCP server management — discover, connect, and use Model Context Protocol tools
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the MCP Tools skill
What this skill tells your AI
The instructions your AI receives, as published by neuron-mr-white/unipi in packages/mcp/skills/mcp/SKILL.md and read by ahel’s review.
MCP (Model Context Protocol) servers provide external tools that extend pi's capabilities.
Tools from MCP servers are named {serverName}__{toolName} — e.g., github__search_code.
Adding Servers
Use /unipi:mcp-add to browse the catalog of 7,800+ MCP servers and add them interactively.
The split-pane overlay lets you:
- Browse: Search the cached server catalog by name, description, or category
- Select: Pick a server to get a pre-filled config template
- Custom: Press
cfor an empty editor to add unlisted servers manually - Save: Press Enter or Ctrl+S to validate and save
Managing Servers
Use /unipi:mcp-settings to manage configured servers:
- Toggle: Press Space to enable/disable a server
- Delete: Press
dthenyto remove a server - Scope: Press
gfor global config,pfor project config - Sync: Press
sto refresh the catalog from GitHub
Quick Status
Use /unipi:mcp-status for a text summary of all servers and their status.
Config Hierarchy
MCP config supports global defaults with project overrides:
~/.unipi/config/mcp/ ← Global defaults (shared across all projects)
{project}/.unipi/config/mcp/ ← Project overrides (when present)
Merge rules:
- Server only in global → used normally
- Server only in project → used normally
- Server in both → project wins entirely
enabled: falsein project metadata → disabled even if defined globally
Config Files
Each level has three files:
mcp-config.json— Server definitions (standard MCP format, portable)config.json— Metadata (enabled/disabled, sync prefs)auth.json— Sensitive env vars (optional, chmod 600)
Example mcp-config.json
{
"mcpServers": {
"github": {
"command": "docker",
"args": ["run", "-i", "--rm", "-e", "GITHUB_PERSONAL_ACCESS_TOKEN", "ghcr.io/github/github-mcp-server"],
"env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxx" }
},
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/home/pi/projects"]
}
}
}
Using MCP Tools
Once an MCP server is running, its tools are available as pi tools.
They're named with the pattern {serverName}__{toolName}.
For example, if you add the GitHub MCP server, you'll have tools like:
github__search_codegithub__create_issuegithub__list_pull_requests
You can call these tools directly in conversations.
Troubleshooting
Server won't start
- Check
/unipi:mcp-statusfor error details - Verify the command exists:
which npxorwhich docker - Check that required env vars are set in the config
Tools not appearing
- Ensure the server status shows "running"
- Check if the server supports the MCP tools protocol
- Try restarting pi after adding a new server
Config issues
- Validate JSON syntax in your config files
- Check file permissions (mcp-config.json should be readable)
- Use
/unipi:mcp-settingsto view current configuration
Catalog sync issues
- Run
/unipi:mcp-syncto force a refresh - Check network connectivity to GitHub
- The seed catalog (49 servers) is available offline as fallback
Signals
- GitHub stars
- 64
- Forks
- 15
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
mcp-neuron-mr-white- Source
- github.com/neuron-mr-white/unipi