Usage scenarios
Worked examples of using the Eventum MCP server — build a generator from a description, manage live generators, and diagnose a failing run.
Each scenario below is a task for an agent you have connected, ordered from the most common to the more specialized. The prompts are illustrative; phrase your own requests freely.
Build a generator from a description
The primary use case: you describe the data you want, and the agent assembles a generator that produces it.
Build a generator that emits one JSON web access-log event per second: a random client IP, an HTTP method, a request path drawn from a small sample, a status code, and a response size. Then preview ten events.
The agent discovers the right plugins, drafts the files, and checks each step against Eventum until the generator loads cleanly and the preview matches your request. The result is a saved, validated generator you can run with eventum run or eventum generate. The create_generator prompt guides the agent through this.
Ask to preview at any point before saving — the schedule first, then the rendered events. Nothing is written or started until you approve it.
Manage live generators over HTTP
With the HTTP transport enabled and writes allowed, the agent can operate the generators on a running server, not just author them.
What's running right now? Start the
web-access-loggenerator and tell me when it's producing events.
The agent can list the running generators and start or stop them. It can also register one it just authored, then start it — and the server restores it after a restart. Every operation is protected by your server's authentication. The live_ops prompt walks through these operations.
Diagnose a failing generator
When a live run fails, the agent can find the cause without you opening a log file.
The
payments-streamgenerator stopped — what went wrong?
The agent checks the generator's status, sees that the run ended in failure, and reads its recent logs to find the cause — a bad configuration value, an unreachable output. It then proposes a fix, rewrites the file, and restarts the generator.