Connect · reference

Advanced connect & reference

The one-command path is on the connect page. This is everything underneath it: endpoints, the ops tool profile, generating a config from inside your agent, and long-lived tokens for daemons and CI.

Endpoints

https://mcp.flukebase.me/mcp default

Memory, sessions, tasks, projects, content, browser + more — the everyday builder surface (~748 tools).

https://mcp.flukebase.me/mcp/ops power user

Deploy, infra, CI, runners, payments, workflow. Same connect command — just swap the URL. Bind a token to one endpoint; it won't work on the other.

Already connected? Generate a config

Once your agent is connected, it can write its own config or list what it can reach:

  • flukebase_project_mcp_config — lists available tools by category and can emit ready-to-paste claude mcp add commands.
  • flukebase_project_mcp_connect — writes an .mcp.json to your project with HTTP transport + bearer auth.

Long-lived tokens (daemons & CI)

Interactive OAuth is ideal for a laptop. For a background daemon or CI job that can't open a browser, mint a long-lived token with flukebase_create_api_token and pass it as the bearer:

claude mcp add --transport http flukebase \
  --header "Authorization: Bearer $MCP_AUTH_TOKEN" \
  https://mcp.flukebase.me/mcp

Treat the token like a password — keep it in an env var / secret store, never commit it.

How auth works

FlukeBase is an OAuth 2.1 authorization server with Dynamic Client Registration (RFC 7591) and PKCE (RFC 7636). On first connect your harness registers itself, opens a browser to sign in, and — if you don't have an account yet — one is provisioned on the spot. Access tokens are short-lived and refresh automatically; each token is bound to the endpoint it was issued for.