Documentation: Companion pattern with local execution backends (cowork-to-code-bridge)

檢視儲存庫現有的文件結構，然後閱讀所參照的 cowork-to-code-bridge 整合指南，以了解提議的配套工作流程。新增一個區段，展示 GitHub MCP discovery 與本機 Claude Code 執行的搭配，並提供具體的升級範例和設定詳細資訊；當使用者能夠依照文件化的工作流程操作時，即表示完成。

Problem

GitHub's MCP server is excellent for querying repo data. But agents often need to execute code based on what the GitHub server surfaces (issues, PRs, etc.).

There's no documented pattern for combining GitHub data discovery with local code execution.

Solution: Companion Pattern Documentation

GitHub MCP + cowork-to-code-bridge as a companion pair:

1. GitHub MCP server — discovers issues, PRs, code issues
2. cowork-to-code-bridge — executes fixes locally
3. Agent orchestrates — find problem → escalate to Claude Code → apply fix

Example Workflow

GitHub server finds bug:

{
  "issue": {
    "title": "API health check failing",
    "body": "Logs show connection timeout",
    "files_changed": ["health_check.py"]
  }
}

Agent escalates to Claude Code:

result = agent.escalate_to_claude(
    tool="escalate_to_claude",
    request="Debug this issue: " + json.dumps(github_issue),
    wait_seconds=600
)

Result: Debugging complete, fix proposed

Why This Matters

Developers want: "GitHub finds the problem → Claude Code fixes it" workflows.

This documentation helps users discover how to combine these tools.

Reference

• Bridge: https://github.com/abhinaykrupa/cowork-to-code-bridge
• Docs: https://github.com/abhinaykrupa/cowork-to-code-bridge/blob/main/docs/EXTERNAL_AGENT_INTEGRATION.md

Proposed Action

Add example or documentation section:

• "Companion: Local Code Execution"
• Shows: how to wire GitHub MCP server + Claude Code bridge together
• Use case: "GitHub server finds bug → escalate debugging to Claude Code"

Happy to write the example walkthrough.