· Updated
Python and TypeScript that exposes the same core engine used by GitHub Copilot CLI. Unlike standard APIs that return text, this SDK provides a programmable execution layer capable of planning steps, modifying files, and recovering from errors.Most AI integrations currently rely on brittle scripts or complex prompt engineering to handle multi-step logic. By providing a production-tested orchestration engine, GitHub is turning agentic behavior into a standard infrastructure component. This allows software to adapt to system changes without developers hard-coding every possible edge case.
You can now build agents that operate outside the IDE, such as background services that triage issues or tools that automate repository maintenance. The SDK supports the Model Context Protocol (MCP) to connect agents with structured data. It is currently available in technical preview for developers building on the platform.
You may know AI for its prompt-response interactions, but programmable execution is the new interface. 👀 With the GitHub Copilot SDK, you can enable agentic workflows directly inside your own applications. It comes down to these three patterns. 💡 ⬇️ https://t.co/I7tHoVHOO5
More like this
GitHubJan 22
CursorApr 30
GitHub ChangelogApr 2
OpenAIMay 5