Senior Engineer · Posted 17 days ago
Our customer data is in Postgres, not in Jarvis. When a customer asks 'what's the status of my order 123', Jarvis needs to query our orders table. Is this possible? Do I need to proxy through an API?
Senior Engineer · Posted 17 days ago
Our customer data is in Postgres, not in Jarvis. When a customer asks 'what's the status of my order 123', Jarvis needs to query our orders table. Is this possible? Do I need to proxy through an API?
Support Engineer · DeskClone AI
Two ways depending on how much trust you want to give Jarvis:
Path A - Direct read-only DB connection (Business plan only)
Settings > Integrations > Database > Add Connection. Provide read-only credentials, allowlist of tables, allowlist of columns. Jarvis gets a query_database tool scoped to exactly what you allowlisted. Queries are logged and auditable.
Path B - Custom API proxy (all plans)
You expose an endpoint like GET /orders/:id on your backend. Add it to Jarvis as a custom API tool. Jarvis calls it with the order ID, your API returns JSON, Jarvis uses the result.
Path B is what most teams start with. It gives you full control over what Jarvis can see (filter fields, mask PII before returning), and you can add caching / rate limits at your API layer.
Senior Engineer
Going with Path B. Built a /orders/:id endpoint in an afternoon, wired it up as a custom tool. Jarvis started using it on the next session. Worth noting for others: the API spec uploader accepts OpenAPI 3.0, so I dropped our swagger.json and it auto-generated the tool definitions.