Agent Playground is liveTry it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
ToolExperimental

helmor

by dohooo

Local-first TypeScript workbench for developing and replaying multi-agent systems

TypeScript
Updated Jul 10, 2026
Share:
1.3k
Stars
115
Forks

View on GitHub

What It Does

Provides a local-first workbench for building and iterating on multi-agent systems in TypeScript, following the Agentic RAG Pattern. It runs agents and conversations locally, captures interactions, and gives developers a sandboxed environment compliant with the Open Agent Specification for rapid prototyping and replaying scenarios. Notable features include git-worktree-friendly project layout and tooling focused on developer workflows for agent collaboration and debugging.

The Value Proposition

As agents are composed and delegated more often, testing interaction patterns locally before deployment becomes essential for trust. Mutual Verification Pattern helps establish a trusted agent track record by enabling reproducible runs and inspection of failures, which this workbench makes easy to reproduce.

Best For

Developers and teams building and debugging multi-agent workflows who need a local sandbox to reproduce interactions and failure modes, including clear handoffs in multi-agent workflows Handoff Pattern.

Applications

  • Reproduce and debug agent-to-agent conversations locally to find coordination bugs
  • Iterate on agent delegation patterns and replay scenarios for regression testing
  • Maintain a local record of runs to build informal agent track records and surface recurring failures
Topics
agent-orchestrationaiclaude-codecodexcoding-agentsdeveloper-toolsgit-worktreelocal-firstmacosmulti-agent+2 more
Similar Tools
autogencrewai
Keywords
multi-agent trustmulti-agent orchestrationagent track record