MCP connector

Kabaido runs a Model Context Protocol server at /api/mcp using streamable HTTP. It gives AI assistants scoped, read-mostly access to your organisation plus the ability to raise requests. Available on plans that include API access.

Connect from Claude

1. In Claude open Settings, then Connectors, then Add custom connector.
2. Paste https://kabaido.ai/api/mcp and continue.
3. Claude sends you to Kabaido to sign in. Pick the organisation and the scopes to grant, then Allow access.

The connection uses OAuth with PKCE; no key is pasted anywhere. You can revoke it at any time from Settings, then Integrations, under MCP.

Connect from Claude Code

claude mcp add --transport http kabaido https://kabaido.ai/api/mcp \
 --header "Authorization: Bearer kbd_your_key_here"

Connect from Cursor and other clients

{
 "mcpServers": {
 "kabaido": {
 "url": "https://kabaido.ai/api/mcp",
 "headers": { "Authorization": "Bearer kbd_your_key_here" }
 }
 }
}

Tools

A tool call outside the connection's scopes returns a clear error rather than data. Scopes are fixed when the connection is approved; reconnect to change them.

Usage and credits

Reading data over MCP costs nothing, like browsing the app. Actions that create work differ by credential. Through an API key, create_request is free, the same as the REST API. Through an OAuth connector such as the Claude connector, create_request charges credits at twice the standard line weight, because the assistant's reasoning runs on your own AI account rather than the Kabaido AEI; the connector deposits finished work, so the platform meters it at 2x in place of the per-line AEI charges it bypasses. The charge appears in your credit ledger with the connector recorded, and a connection without enough credits receives a clear error instead of silently failing.