Cowork to Code Bridge is an open-source orchestration harness developed by Abhinay Krupa. It connects Claude interfaces (such as Claude Cowork or desktop-based Claude workflows) with local developer execution environments running Claude Code. The bridge enables remote or sandboxed Claude environments to dispatch coding tasks and command executions directly to a local development machine. Operationally, the harness runs a local background daemon that accepts execution requests through a client library (`bridge_client.py`) and a Claude plugin integration (`.claude-plugin`). It supports synchronous execution via remote calls as well as an asynchronous queue mechanism (`queue_task` and `poll_task_result`). This async pattern allows web or sandbox sessions with short command timeouts (such as 45-second bash execution limits) to trigger long-running Claude Code sessions locally and poll for completed results incrementally. What differentiates Cowork to Code Bridge is its focus on bridging context boundaries between Anthropic's interactive environments and local CLI agent execution. Rather than reinventing agentic coding logic, it provides the bridge daemon, task queue, and client interfaces needed to let conversational Claude instances securely delegate filesystem and terminal tasks to local Claude Code instances. The project is open-source on GitHub and distributed as a Python package on PyPI, making it free to use alongside user-provided Anthropic Claude API credentials or Claude Code subscriptions.
Tags: ai agent, cli, coding agent, developer tools, OpenSource
queue_task() and polls with poll_task_result() to bypass sandbox execution timeouts. - Claude Plugin Integration — Provides a .claude-plugin configuration for direct discovery and invocation from compatible Claude interfaces. - Local Daemon Architecture — Runs a background execution bridge on the local machine managing agent execution. - Claude Code Orchestration — Delegates tasks directly to local Claude Code runs without manual terminal switching. - Synchronous & Asynchronous Execution Modes — Supports both immediate blocking execution via remote calls and non-blocking background task polling.This tool is intended for developers using Anthropic Claude interfaces alongside Claude Code who want seamless delegation between interactive chat environments and local terminal execution. It is not designed for teams looking for standalone, non-Anthropic AI coding engines or all-in-one proprietary IDE replacements.
bash pip install cowork-to-code-bridge Alternatively, clone the repository and install from source: bash git clone https://github.com/abhinaykrupa/cowork-to-code-bridge.git cd cowork-to-code-bridge pip install -e .bash # Start the bridge daemon on your local workstation cowork-bridge start Verify that the bridge root directory is properly configured and accessible.python from cowork_to_code_bridge import queue_task, poll_task_result # Queue a task for local Claude Code execution task = queue_task('scripts/run_claude.sh', args=['Refactor auth middleware to use JWT verification'], timeout=600) task_id = task['task_id'] print(f'Task queued: {task_id}') # Poll for completion status result = poll_task_result(task_id) print(result)Part of CLI Coding Agents
Categories: AI Coding Agents, Code Assistants, Developer Tools