I built PolyClaw, an autonomous agent for the PolyMCP ecosystem inspired by OpenClaw.

PolyClaw doesn’t just call tools. It plans, executes, adapts — and even creates MCP servers when needed.

It’s designed for real-world, multi-step production workflows where an agent must: • Orchestrate multiple tools • Spin up infrastructure dynamically • Recover from failures • Deliver complete, end-to-end results

⸻

What PolyClaw Does • Decomposes complex tasks into executable steps • Dynamically selects and orchestrates MCP tools • Spins up or connects to MCP servers on demand • Adapts if execution fails or context changes • Validates outputs before proceeding • Runs Docker-first for isolation and safety • Built with Python + TypeScript

PolyClaw is not just a tool-caller — it’s an infrastructure-aware agent.

⸻

Run PolyClaw (via PolyMCP CLI)

polymcp agent run \ --type polyclaw \ --query "Build a sales reporting pipeline and test it end-to-end" \ --model minimax-m2.5:cloud \ --verbose

What happens behind the scenes: 1. The task is decomposed into structured steps 2. Required MCP tools are identified 3. MCP servers are started or connected 4. Steps execute (sequentially or in parallel) 5. Outputs are validated 6. Failures trigger adaptive replanning 7. A complete, end-to-end result is returned

All containerized. All isolated.

⸻

Why This Matters

Most AI agents today: • Call tools statically • Assume infrastructure already exists • Break on multi-step failures

PolyClaw instead: • Builds the infrastructure it needs • Orchestrates across multiple MCP servers • Handles retries and adaptive planning • Is safe to run in Dockerized environments

This makes it viable for: • Enterprise workflows • DevOps automation • Data pipelines • Internal tooling orchestration • Complex multi-tool reasoning tasks

PolyClaw turns PolyMCP from simple tool exposure into a fully autonomous orchestration layer.

Repo: https://github.com/poly-mcp/PolyMCP

Happy to answer questions.