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
ToolProduction Ready

claude_codex_bridge

by SeemSeam

Real-time multi-LLM collaboration with persistent, low-cost shared context

Python
Updated Jul 27, 2026
Share:
3.3k
Stars
323
Forks
14
Commits/Month

View on GitHub

Summary

Enable real-time collaboration between Claude, Codex, and Gemini while keeping context persistent and token overhead low. Routes messages between models, maintains a compact shared context (real-time collaboration and context-delta packing) and provides a CLI plus Python primitives for live coordination. Includes replayable logs and minimal-state context syncing to reduce cost and make conversations reproducible.

Why It Matters

As agents delegate and combine capabilities, keeping a compact, auditable conversation history becomes essential for trust and debugging. Persistent, low-overhead context and replayable interaction logs let teams build agent track records and inspect failure modes without huge token costs. This repo makes multi-model collaboration practical and easier to evaluate for agent-to-agent behavior and reliability, supporting auditable conversation history.

Ideal For

Teams wiring multiple LLMs (Claude, Codex, Gemini) together for live collaboration or prototyping multi-agent flows with reproducible logs. Consider adopting patterns like planning to coordinate steps in complex tasks and market-based coordination for scalable agent interaction.

Applications

  • Prototype multi-model pipelines that delegate subtasking to specialized LLMs
  • Record and replay agent interactions to build an agent track record
  • Reduce token costs by syncing compact shared context across agents
Works With
openaianthropicgooglepython
Topics
ai-collaborationai-toolsclaude-codeclicodexcross-platformgeminillmmulti-agent-systemspython+1 more
Similar Tools
autogencrewai
Keywords
multi-agent orchestrationmulti-agent trustagent-to-agent evaluationpython