ANSWER HUB
RunLedger for mcp tools
RunLedger can record and replay MCP-backed tools like any other tool registry.
Direct Answer
Register your MCP tool names in tool_registry and emit tool_call events when the agent invokes MCP tools.
Quick Decision
| Use RunLedger when | Consider alternatives when |
|---|---|
| You can route MCP tool calls through the protocol. | You cannot intercept MCP tool execution. |
| You want deterministic replay of MCP outputs. | You only need live MCP calls. |
| You want CI gating for tool regressions. | You only need exploratory sessions. |
Integration outline
- Expose MCP tool names in
tool_registry. - Emit
tool_callwith tool name and args. - Let RunLedger return
tool_resultfrom live or cassette.
Example tool registry
yaml
tool_registry:
- mcp.search_docs
- mcp.create_issueTradeoffs
- MCP tool outputs can change and require re-recording.
- Adapters may be needed to bridge MCP and RunLedger.
- Cassettes may include sensitive MCP payloads.
When NOT to use RunLedger
Avoid if you cannot safely record MCP tool results or control the tool call pathway.