GitHub has open sourced the internal agent runtime that powers the GitHub Copilot CLI and exposes it as a programmable SDK. GitHub copilot-sdkNow in Technical Preview, lets you embed the same agentic execution loop into any application so the agent can plan, launch tools, edit files, and run commands as part of your own workflow.
What does GitHub Copilot SDK provide
GitHub Copilot-SDK is a multi-platform SDK for integrating the GitHub Copilot agent into applications and services. It provides programmatic access to the execution loop that already powers the GitHub Copilot CLI. Instead of creating your own planner and tool loop for each project, you connect your logic to this existing runtime and treat it as an execution platform.
GitHub Copilot-SDK exposes the same production test runtime used by the Copilot CLI, with support for multi-model operations, multi-step planning, tools, Model Context Protocol (MCP) integration, authentication, and streaming. This gives you the same agent behavior that CoPilot uses in the terminal, but callable from your own code.
Agentic execution loop as runtime primitive
The main abstraction is the agentic execution loop. In the CoPilot CLI and SDK, interactions are not separate prompts. The agent maintains state at every turn, chooses plans, calls tools, executes commands, reads results and repeats these steps until it reaches the goal you provide.
The GitHub team describes common problems when you implement this loop yourself. You need to manage the context in multiple ways, organize external tools and commands, route calls to models, integrate MCP servers, and think through Permit Developer, you focus on defining domain specific tools, describing functions, and defining what agents can do.
Supported languages ​​and core API
CoPilot-SDK is available in 4 languages ​​in this technical preview:
- Node.js and TypeScript, via packages
@github/copilot-cli-sdk - Python, via package
copilot - go through modules
github.com/github/copilot-cli-sdk-go - .NET, via packages
GitHub.Copilot.SDK
All SDKs expose a consistent API surface. According to the changelog, each language binding supports multi-turn conversations with session history, custom tool execution, and programmatic control over the client and session lifecycle.
Integration with tools, MCP servers and existing systems
A key feature of the CoPilot agent is device execution. Through the SDK you can register custom tools that the model can call during interactions. Copilot-CLI already exposes custom tool definitions and full MCP server integration, and the SDK reuses that capability.
MCP gives agents a standard protocol to connect to external systems such as internal APIs, document stores, or operations tools. When you integrate an MCP server, the CoPilot agent can discover and call its operations in a structured manner with consistent metadata instead of ad-hoc prompt engineering.
The pattern is straightforward. You define a tool with clear schema and effects, you expose it via the SDK, and Copilot Planner decides when and how to call it as part of a multi step plan.
Authentication, Subscription and Streaming
The SDK integrates with GitHub authentication and Copilot subscription. You can either use an existing GitHub Copilot subscription or bring your own key when configuring the SDK. This is important when you embed the agent in an enterprise environment where identity and access control is already standardized around GitHub.
Streaming is part of the contract. Copilot-CLI already supports real-time streaming in the terminal, and the SDK exposes streaming so that applications can receive responses sequentially. This allows you to create a user interface that updates progressively as the agent causes and executes, without waiting for complete completion.
Link to GitHub Copilot-CLI
The SDK is not a separate agent implementation. This is a layer on top of the existing Copilot CLI execution loop. It is a way to reuse the CLI’s planning, tool usage, and multi-turn execution behavior in any environment.
Copilot-CLI is self-evolving. Recent updates include persistent memory, infinite session and context compilation, support for tracing and planning workflows with model selection per step, custom agents and agent skills, full MCP support, and asynchronous task delegation. The SDK benefits from this functionality, as it exposes the same behavior through language specific libraries.
key takeaways
- The GitHub Copilot-SDK exposes the same agentic execution loop that powers the GitHub Copilot CLI, so applications can call a production tested planner that runs multi step workflows with tools and commands.
- The SDK is available for Node.js, Python, Go, and .NET, and each language binding provides a similar abstraction around clients and sessions that manage multi-turn conversations and tool usage.
- Developers define domain specific tools and model context protocol servers, then register them through the SDK, and the Copilot agent decides when and how to call them as part of the scheme.
- The runtime integrates with GitHub authentication and Copilot subscriptions, supports multiple AI models like a GPT based backend, and exposes real-time streaming so applications can render partial responses sequentially.
check it out GitHub page. Also, feel free to follow us Twitter And don’t forget to join us 100k+ ml subreddit and subscribe our newsletter. wait! Are you on Telegram? Now you can also connect with us on Telegram.
Michael Sutter is a data science professional and holds a Master of Science in Data Science from the University of Padova. With a solid foundation in statistical analysis, machine learning, and data engineering, Michael excels in transforming complex datasets into actionable insights.
