What it does
Agent Replay Debugger MCP records every step an AI agent takes and can replay those steps deterministically, working as a step debugger for agent runs. It also produces signed audit evidence of the recorded steps. Tools named in the publisher's examples include start_recording, record_step and replay_step.
Developers would reach for it to step through how an agent behaved or to reproduce a run. The publisher's documentation states that the server addresses EU AI Act articles 9, 13, 15, 26 and 50, where applicable.
Running it means adding the server to an MCP client configuration, such as the claude_desktop_config.json file used by Claude Desktop, or installing the Python package and launching it over stdio transport. The project is released under the MIT license by MEOK AI Labs.
Limitations and requirements
Requires an MCP client configured to launch the server over stdio transport. The server is installed as a Python package.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
io.github.CSOAI-ORG/agent-replay-debugger-mcp
Agent Replay Debugger MCP — record every agent step + deterministic replay. Step-debugger for
v1.0.7As shown on registry.modelcontextprotocol.io when we last read it