Base URL
https://mcp.crevio.co/mcp/server-card, listed in the AI catalog at https://crevio.co/.well-known/ai-catalog.json.
Authentication
Every request carries a Bearer token in theAuthorization header — the same API tokens from your Developer settings.
401 with a WWW-Authenticate header pointing at /.well-known/oauth-protected-resource, so clients that support MCP’s OAuth flow (Claude’s connector directory, ChatGPT) can sign in interactively instead of pasting a token.
Claude Code
Claude Desktop
Add this toclaude_desktop_config.json and restart Claude Desktop:
“Ask Crevio to create a 20% discount code called SUMMER20 that expires next month”
“Have Crevio draft and schedule three posts about my new course, then tell me what it scheduled”
“Who are my top 5 customers by order count?” (the agent may answer this directly with api_execute)
Cursor
Add to~/.cursor/mcp.json (or Settings → MCP Servers → Add, type Streamable HTTP):
Other clients
Most MCP clients accept the same JSON shape: aurl plus an Authorization header. For clients that drive OAuth themselves, point them at https://mcp.crevio.co/mcp and let the 401 challenge start the sign-in.
Verify the connection
Callwhoami. It takes no arguments, has no side effects, and returns the account and user behind the credential, the plan and credit balance, rate limits, wait limits, and the full tool list.
whoami comes back, the connection is good. If it doesn’t:
- 401 — the token is missing, mistyped, expired, or revoked. Mint a new one in Settings → Developers.
- Tools missing in the client — most clients cache
tools/list; reconnect or restart the client. - 429 — the token is over its 600-requests-per-minute budget (shared with the REST API); back off until
X-RateLimit-Reset.
First delegation
result. If it comes back as wait_timed_out, the job is still going — call wait_for_run with the run’s id.
See the Overview for the full tool reference — delegation (
ask_crevio, start_chat, wait_for_run, send_message, …) and direct API access (api_search, api_execute).
