Trace2MCP: Compile Agent Traces into Verified MCP Workflows
War_Enterprise · reddit · 2026-08-25
Trace2MCP is an open-source Python project that records a successful agent execution and compiles its tool-call trace into a deterministic workflow, preventing agents from rediscovering the same tool procedure on every run.
Trace2MCP 0.2.0 capabilities include:
- Infer dependencies: Infer dependencies between observed tool calls.
- Build DAG: Build a parallelizable Directed Acyclic Graph.
- Offline verification: Verify references, integrity hashes, and safety policies offline.
- Deterministic replay: Perform deterministic frozen replay without invoking tools.
- Generate project: Generate an MCP-ready Python project with typed inputs and inert handler stubs.
- Secure defaults: Reject unknown and destructive operations by default.
- Review contracts: Require reviewed contracts and explicit approval for consequential side effects.
Includes a quick start demo requiring no model or API key. The author notes this is an alpha research project.
More from coding & agent
- Livestream: Running ComfyUI Locally via MCP and Hardware Optimization — MiniMax_AI · 2026-08-25
- Headlong experiments with persistent agency via exponential backoff — lateinteraction · 2026-08-25
- Grok Build VS Code Extension Released with Remote Control — PawelHuryn · 2026-08-25
- Event: Building AI agents with retrieval backend from scratch — hugobowne · 2026-08-25
- Alchemy AWS Emulator Patch Fixes Gaps in Floci — samgoodwin89 · 2026-08-25
- Google ADK Introduces Live Evaluation for Voice-Based Agents — rseroter · 2026-08-25