One-click install
Add the Foldspace connector to Claude →
Connector URL: https://api.foldspace.ai/mcp
Get started
Connect Foldspace to the AI coding tool you already use. Add the Foldspace MCP server to your editor, and your assistant can browse your actions and implement them in your code.
Using Claude on the web or desktop? Add Foldspace as a custom connector. The link pre-fills the name and URL, so all you do is confirm and authenticate:
One-click install
Add the Foldspace connector to Claude →
Connector URL: https://api.foldspace.ai/mcp
In Claude Code, add the same server from the terminal instead:
claude mcp add --transport http foldspace https://api.foldspace.ai/mcpThe Model Context Protocol (MCP) is a standard for giving AI models real-time, external context. Foldspace exposes two MCP servers. Add either or both to your editor:
| Docs MCP | Product MCP | |
|---|---|---|
| Endpoint | https://docs.foldspace.ai/mcp | https://api.foldspace.ai/mcp |
| Auth | Public (no key) | OAuth (sign in from your client) |
| Gives your assistant | Semantic search + read access to these docs (search_docs, fetch_page) | Access to your account’s AI Actions: list them, read schemas, generate handlers |
| Best for | Any developer: ask questions about Foldspace while you code | Building on your own Foldspace agent |
Start with the Docs MCP: there’s nothing to configure but a URL. Add the Product MCP when you want your assistant to work with your own AI Actions.
The Docs MCP is read-only and needs no API key. Point any MCP client at https://docs.foldspace.ai/mcp. It exposes two tools, search_docs (semantic search over the docs) and fetch_page (fetch a page as Markdown), so you can ask your assistant things like “how do I configure JWT auth in Foldspace?” and it answers from the current docs.
Claude Code
claude mcp add --transport http foldspace-docs https://docs.foldspace.ai/mcpCursor: Settings → Tools & Integrations → New MCP Server:
{ "mcpServers": { "foldspace-docs": { "url": "https://docs.foldspace.ai/mcp" } }}Windsurf: Cascade → hammer icon → Configure → View raw config:
{ "mcpServers": { "foldspace-docs": { "serverUrl": "https://docs.foldspace.ai/mcp" } }}VS Code (GitHub Copilot): in .vscode/mcp.json:
{ "servers": { "foldspace-docs": { "type": "http", "url": "https://docs.foldspace.ai/mcp" } }}The Product MCP exposes your Foldspace AI Actions as callable tools. Connecting your editor to it lets your AI assistant:
It’s a hosted HTTP server at https://api.foldspace.ai/mcp. There’s no API key to paste and
nothing to install. The first time your client connects, Foldspace opens a browser window
for you to sign in, and the client stores the resulting OAuth token itself.
Use the one-click connector link above, or in Claude Code:
claude mcp add --transport http foldspace https://api.foldspace.ai/mcpThen run /mcp and pick foldspace to complete the OAuth sign-in.
Settings → Tools & Integrations → New MCP Server:
{ "mcpServers": { "foldspace": { "url": "https://api.foldspace.ai/mcp" } }}Cursor prompts you to authenticate the first time it connects.
Cascade → hammer icon → Configure → View raw config:
{ "mcpServers": { "foldspace": { "serverUrl": "https://api.foldspace.ai/mcp" } }}Open Settings, search for “MCP”, and make sure GitHub > Copilot > Chat: MCP is enabled.
Then create .vscode/mcp.json at your workspace root:
{ "servers": { "foldspace": { "type": "http", "url": "https://api.foldspace.ai/mcp" } }}Reload the window, then use # in the Copilot Chat panel to reference the Foldspace tools.
Antigravity has no one-click install yet. Open Settings → Customizations → Add MCP+, or
edit ~/.gemini/config/mcp_config.json, and add:
{ "mcpServers": { "foldspace": { "url": "https://api.foldspace.ai/mcp" } }}Once the connector is authorized, the Foldspace tools show up in your client. Here’s how it looks in Claude’s connector settings:
https://api.foldspace.ai/mcp discover_actions foldspace_overview generate_action_handler get_action get_action_schema get_action_version get_agent get_agent_install_snippet Tools are grouped by what they touch, so you can let read-only calls run freely and keep approval on anything that writes.