Local vs Remote MCP — which to use
Key capabilities
Security
- API key auth: Requires
YUNO_PUBLIC_API_KEY,YUNO_PRIVATE_SECRET_KEY, andYUNO_ACCOUNT_CODE - IP validation: Sessions are bound to the client IP that initiated them. Requests from a different IP within the same session are rejected.
- Rate limiting: 15 requests per minute per session
- Session timeouts: 30 minutes idle; 6 hours absolute
Protocols
- Streamable HTTP transport for bidirectional communication
- JSON‑RPC 2.0 for standardized request/response
Session management
- Automatic creation, refresh, and cleanup
- Scheduled cleanup every 5 minutes
- Redis backend for persistent session and rate‑limit storage
When to use the remote server
Use the Remote MCP Server when you want centralized security, observability, and policy enforcement (IP binding, rate limits, TTLs), or when teams and tools should connect without distributing raw API credentials locally. For rapid prototyping on a developer machine, the local MCP (CLI‑launched) may be sufficient.Endpoints
Use the matching credentials for each environment — sandbox keys against the sandbox URL, production keys against the production URL.
Connecting from MCP‑compatible clients
The Remote MCP Server exposes an HTTP transport endpoint. Configure your MCP client to connect to your provisioned URL and pass the required headers.- Claude Desktop / Cursor
- cURL probe
Add to your client’s MCP config:Restart the client. Yuno tools should appear in the agent’s tool palette.
Your
PUBLIC_API_KEY, PRIVATE_SECRET_KEY, and ACCOUNT_CODE can all be obtained from the Yuno dashboard.Available tools
The Remote MCP Server exposes the same tool catalog as the local MCP — every Yuno API endpoint is mapped to a callable tool (customers, payments, payment methods, checkout sessions, payment links, subscriptions, recipients, installment plans, and more). For the full list, see the tool catalog in Building AI Integrations.Errors and troubleshooting
Operational notes
- Logging: Yuno logs the tool name, request timestamp, source IP, and account code per call. Request bodies are not retained beyond what the underlying API logs.
- Concurrency: 15 req/min per session. To run multiple agents in parallel, open multiple sessions (each gets its own quota).
- Credential rotation: Rotate via Dashboard → Settings → Developers (Credentials). Existing sessions continue with the prior credential until they expire; new sessions must use the new credential.
Related information
- Building AI Integrations with Yuno’s LLMs and MCP — local MCP setup and full tool catalog
- LLM Agent Instructions — canonical doc URLs and SDK vs Direct Flow ordering tuned for AI agents
- Aida AI Agent — Yuno’s hosted dashboard agent (different product, same underlying APIs)