MCP Servers
MCP (Model Context Protocol) servers extend PandaOS's capabilities by giving it access to external tools and data sources. PandaOS lets you configure MCP servers both globally and per-project.
What Are MCP Servers?
MCP servers are programs that expose tools to PandaOS. For example:
- Filesystem MCP, gives PandaOS direct file system access
- GitHub MCP, lets PandaOS interact with GitHub issues, PRs, and repos
- Database MCP, lets PandaOS query databases
- Browser MCP, PandaOS's built-in server that lets PandaOS interact with your browser preview
When an MCP server is connected, its tools appear automatically in PandaOS's available tool list.
Opening the Integrations page
MCP servers live on the Integrations page in the sidebar.
- Click Integrations in the sidebar
- Browse the catalog, or add a custom server
- Choose whether the server applies to every project or only the current one
Adding an MCP Server
Paste a JSON configuration or a claude mcp add command into the input area. PandaOS auto-parses the input and shows a preview of the detected servers.
Supported input formats:
- CLI command:
claude mcp add --transport stdio my-server -- npx my-mcp-server - JSON config:
{ "mcpServers": { "my-server": { "command": "npx", "args": ["my-mcp-server"] } } } - Bare config:
{ "command": "npx", "args": ["my-mcp-server"] }(you'll be asked for a name)
If you're adding to a project, choose the scope:
- Project-shared (
.mcp.json), Committed to git, shared with your team - Project-local (
~/.claude.json), Personal project overrides, not committed - Global (
~/.claude.json), Available across all projects
Click "Add MCP Server" to save. Quick-links below the input let you open ~/.claude.json or .mcp.json directly for manual editing.
Server List
Each configured server shows:
- Status indicator, green (enabled), gray (disabled), red (auth error), yellow (authenticating)
- Type icon, globe for remote (HTTP/SSE) servers, terminal for local (stdio) servers
- Server name and subtitle (URL or command)
- Source badge, where the config lives (global, project-shared, project-local)
- Inherited badge, for global servers shown in project view
Click a server to expand it and see its available tools (tool name and description).
Server Controls
For each server:
| Action | Description |
|---|---|
| Connect | Authenticate a remote server that requires auth |
| Disable | Temporarily disable the server without removing it |
| Enable | Re-enable a disabled server |
| Remove | Delete the server configuration |
Scopes
- Global servers, available in all projects. Stored in
~/.claude.json. - Project-shared servers, only for that project, committed to git. Stored in
.mcp.jsonin the project root. - Project-local servers, only for that project, personal (not committed). Stored in the projects section of
~/.claude.json. - In project view, global servers appear with an "Inherited" badge.
Built-in: Browser MCP
PandaOS automatically manages a Browser MCP server (pandaos-browser) when you open the Browser Preview panel. This server:
- Starts automatically when the browser preview opens
- Stops automatically when the browser preview closes
- Gives PandaOS tools to navigate, click, type, screenshot, and interact with your app
- Registers itself in your PandaOS configuration automatically
You don't need to configure this manually, it's handled by PandaOS.
Common MCP Servers
Here are some popular MCP servers you might want to add:
| Server | Package | What It Does |
|---|---|---|
| Filesystem | @modelcontextprotocol/server-filesystem | Direct filesystem access |
| GitHub | @modelcontextprotocol/server-github | GitHub API access |
| PostgreSQL | @modelcontextprotocol/server-postgres | PostgreSQL queries |
| Brave Search | @modelcontextprotocol/server-brave-search | Web search via Brave |
| Memory | @modelcontextprotocol/server-memory | Persistent knowledge graph |
Deployment
Deploy apps and projects to production with PandaOS. Auto-deploy handles everything automatically, or use your own Vercel integration for full control.
Skills
Skills are custom slash commands that extend PandaOS's capabilities with predefined prompts and instructions. They let you create reusable workflows triggered by a simple `/command`.