Guides
Guides
How to give an AI agent access to a real database without regretting it. Each guide answers the question in its first paragraph and starts with the approach that costs nothing: a read-only role, a view layer, statement logging. The SQL is real and was executed before it was published. Then it says precisely what that approach does not cover, and only after that where PgBeam fits, with our own limits attached.
If your question is about how PgBeam stacks up against a specific alternative rather than about a technique, the comparisons are probably closer.
Updated
How to give Claude Code read-only access to a production database
Create a Postgres role that only holds SELECT, point Claude Code at it, and prove it holds. The full SQL, the MCP config, and the four things a read-only role still does not cover.
Updated
How to mask PII before it reaches an LLM
Mask at the database boundary, not in the prompt. The full SQL for a Postgres masking view, the grants people get wrong, and the hash-equality oracle that defeats it, demonstrated.
Updated
How to audit what an agent ran against your database
Turn on statement logging, give the agent its own role, and read the log. The exact config, real log output, and the four questions a Postgres log cannot answer about an agent.