Mediation MCP tools
Use the Mediation MCP tools to manage meters, runs, and audits safely, with guidance-driven operations and strict rules to protect production pipelines and billing data.
Mediation MCP exposes three tools with a total of 40 operations:
| Tool | Operations | Purpose | Start with |
|---|---|---|---|
manage_meters | 19 | Create, update, validate, and inspect meter definitions, event schemas, event stores, and connections | meter_guidance |
manage_meters_run | 14 | Start and stop runs, check status, review history and output, manage prefetch, query event stores, and ingest events | run_guidance |
manage_meters_audit | 7 | Browse audit trail entries and manage export jobs for offline analysis | audit_guidance |
Each call includes an operation name and its parameters. Do not invent parameters. Use the applicable guidance operation or the tool schema.
You can also ask the MCP server about its capabilities. For example, ask, “What can I do with manage_meters?”
How the agent works
The agent follows rules designed to keep production pipelines and billing data safe:
- Read-only by default. The agent does not run, stop, create, update, or export anything unless you explicitly approve that named action.
- Guidance before action. Before using each tool for the first time, the agent consults that tool guidance operation (meter_guidance, run_guidance, or audit_guidance) rather than guessing parameters.
- Figures come from the API. Run counts, error rates, and timings are reported only as returned by Zuora. The agent does not estimate them.
- Enum values are uppercase. For example, runType is NORMAL or DEBUG, and audit exportType is ERROR or SAMPLE.
- Impact is described before any change. Before a write operation, the agent states the action, its impact, and asks for your confirmation. For example, a force stop on a streaming pipeline can lose in-flight data.