MCP
Install Bindly MCP server with one command: npx bindly-mcp. Connect Claude Code, Claude Desktop, Cursor, and other MCP clients to persistent AI knowledge.
Bindly uses the Model Context Protocol (MCP) to let LLMs access your knowledge. Connect in seconds with one command.
Quick Install
npx bindly-mcp
This auto-detects Claude Code, Claude Desktop, and Cursor, and configures them automatically.
Install for a specific client
npx bindly-mcp --claude-code # Claude Code
npx bindly-mcp --claude-desktop # Claude Desktop
npx bindly-mcp --cursor # Cursor
Claude Web (claude.ai)
You can connect Bindly directly from claude.ai as a custom connector:
- In the chat input, click the + button → Connectors → Manage connectors, or
- Open your profile menu → Customize → Connectors
- Name: Bindly
- Remote MCP server URL: https://mcp.bind.ly/mcp
- OAuth Client ID / Secret (under Advanced settings): leave empty — Bindly does not require them
Once connected, you can use Bindly tools in any Claude conversation on the web.
How it runs: claude.ai connectors run on Anthropic's cloud infrastructure, not on your device. Only publicly reachable HTTPS endpoints work — https://mcp.bind.ly/mcp is fine, but localhost and cloudflared Quick Tunnels are not.
Plan limits (custom connectors are in beta): Free = 1 custom connector, Pro / Max = unlimited, Team / Enterprise = the workspace Owner must enable connectors first.
Manual Setup
Claude Code
claude mcp add --transport http bindly https://mcp.bind.ly/mcp
Claude Desktop
Add to your MCP settings (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"bindly": {
"url": "https://mcp.bind.ly/mcp"
}
}
}
Cursor
Add Bindly as an MCP server in Cursor Settings > MCP, or add to ~/.cursor/mcp.json:
{
"mcpServers": {
"bindly": {
"url": "https://mcp.bind.ly/mcp"
}
}
}
Other MCP Clients
- Streamable HTTP:
https://mcp.bind.ly/mcp
Authentication
Bindly uses progressive authentication — many tools work without login:
Public access (no auth needed)
These tools work immediately after MCP connection:
check_connection— verify connection statusmcp_list_spaces— browse public spacesmcp_search— search public knowledgemcp_get_binding/mcp_get_bindings— read public bindingsmcp_list_bindings— list bindings in public spacesmcp_help— learn how Bindly works
Authenticated access (OAuth auto-triggered)
Call connect() to log in — your MCP client automatically starts OAuth:
What Your LLM Can Do
Once connected, your LLM has access to 29 tools:
- Save knowledge:
mcp_create_binding— save content from conversations - Search:
mcp_search— find knowledge by semantic or keyword search - Retrieve:
mcp_get_binding— get full content with version history - Organize:
mcp_create_set,mcp_add_version_to_set— build collections - Comment:
mcp_add_comment— add notes to Bindings or Versions - Share:
mcp_create_share— create temporary share links
See the MCP Tools Reference for the complete list.
User Keys (Alternative Auth)
For programmatic access or when OAuth isn't available:
userk_* token in your MCP client configurationTroubleshooting
"Connection failed": Check your MCP client supports Streamable HTTP protocol.
"Authentication required": Call connect() — your MCP client will start OAuth automatically. If auto-auth doesn't work, try removing and re-adding Bindly:
- Claude Code:
claude mcp remove bindly && claude mcp add --transport http bindly https://mcp.bind.ly/mcp - Claude Desktop / Cursor: restart the app to re-trigger OAuth
"Permission denied": You may not have access to the requested Space. Use mcp_list_spaces to see your accessible Spaces.
Try It Out
Once connected, try these commands with your AI:
- "List my Bindly spaces"
- "Save this to Bindly"
- "Search my Bindly for [topic]"
Or browse the @bindly Space for guides and documentation.