Search-then-fetch is the whole surface, which makes this cheap in context and easy to reason about: the model finds candidates, then pulls the one record it needs in full. Nothing here writes back to DevRev, so it is safe to attach to an assistant that is answering questions about tickets rather than working them.
A small MCP server over the DevRev APIs. It does two things: run a search against DevRev with a query and a namespace, and fetch everything DevRev holds about one object once you have its id.
- A search across DevRev, scoped by the namespace you pass, so tickets, parts or accounts can be targeted separately — `search`
- The complete record for a single DevRev object by its id — `get_object`
A DevRev account and a personal access token, supplied as `DEVREV_API_KEY`. The package is `devrev-mcp` (0.1.1 in pyproject) and the documented launch is `uvx devrev-mcp`. DevRev's own import path pulls in data from Salesforce, Zendesk and similar sources; whatever you have imported is what search reaches.
One command plus a key — uvx devrev-mcp, then supply credentials
