MCP server overview

The TwinCommit MCP server connects your replication pipelines to AI assistants through the Model Context Protocol. With it, Claude, GitHub Copilot, Cursor and any other MCP client can check lag, read pipeline status, explain errors, verify tables and, if you allow it, start, stop and re-snapshot pipelines.

What you can ask

“Is anything broken in production replication?”
“Why did lag on db2-prod → databricks spike at 14:00?”
“Explain this error and tell me how to fix it.”
“Re-snapshot SALES.ORDERS on the staging pipeline.”
“Check whether the new Oracle source is ready for TwinCommit.”

How it works

 AI client ──stdio/HTTP──▶ @twincommit/mcp ──HTTP──▶ twincommit run (/status, /control)
                                │
                                └──▶ Prometheus (optional, for lag history)

The MCP server is a small Node.js package, @twincommit/mcp, that runs next to your AI client. It talks to the /status endpoint of each running TwinCommit pipeline, and optionally to Prometheus for history. It never connects to your source or target databases directly.

Safe by default

Read-only unless you set TWINCOMMIT_MCP_ALLOW_WRITES=true.
Row values redacted in tail_events unless you set TWINCOMMIT_MCP_ALLOW_ROW_VALUES=true.
Audit log of every tool call.

See MCP security.

Next steps