Convex production data: tables, documents and read-only queries
Polylane investigations can now read the data in your Convex deployments: list the tables, read recent documents and run one-off read-only queries.
What changed
Polylane could already see your Convex projects, deployments, logs and metrics. Agents can now read the data behind them too:
- List the tables of a deployment with their document shapes.
- Read the newest or oldest documents of a table.
- Run a one-off query.
Convex runs it as a sandboxed query function, so it cannot write data or reach the network.
The reads work from every investigation, with no local MCP server. For each read, Polylane creates a deploy key for that one deployment with only the permission the read needs, and deletes it afterwards. Each key also expires after 35 minutes, and its name starts with polylane-ephemeral-. See the Convex integration.
Why it matters
An investigation often comes down to what a record looks like in production. Before this, only someone with access to the Convex dashboard or a local
Convex MCP server could check. Now the agent reads the data itself and keeps what it found in the thread as evidence.
How to use it
- In the
Convex dashboard, open Team Settings → Access Tokens and create a team access token.
- In the Polylane console, open Settings > Clouds, connect
Convex and paste the token.
- Ask in a thread, for example: "Read the newest documents in the
orderstable on production and find the ones without acustomerId."