A coding agent can read your codebase, run your tests, and open a pull request. Point it at your database and the calculus changes: a bad migration isn't something you undo with git, so most teams draw a line the agent isn't allowed to cross.

In this hands-on workshop, you'll wire Tiger MCP into the coding agent you already use and give it a database it's allowed to break. You'll create a Tiger Cloud service, have your agent ingest a real time-series dataset, then fork that service into an instant, isolated copy. From there you'll turn the agent loose on a deliberately bad schema, one with wrong data types, missing indexes, and no hypertable, and watch how it works. It reads EXPLAIN plans, searches Tiger Data documentation through the MCP server itself, and applies the TimescaleDB primitives that fix the problem. You'll measure the result, compare the fork against the original, and throw the fork away.

By the end, you'll have Tiger MCP running against your own agent, a repeatable fork-first workflow for letting agents do real database work, and the guardrails that make it safe to point at something that matters, whether you're tuning a slow dashboard query, planning a migration, or making sense of a schema you inherited.

What you'll learn

• How to connect Tiger MCP to Claude Code, Codex, or any MCP-capable coding agent

• How to fork a Tiger Cloud service so agent-driven changes never touch the original

• How to have an agent diagnose slow queries using EXPLAIN plans and built-in Tiger Data documentation search

• How to apply hypertables and continuous aggregates through an agent, then verify the improvement

Not able to join live? You'll still get the recording if you sign up.