estruyf/copilot-mock-server

copilot-mock-server is a local proxy tool for developers who want to give demos or test tooling that uses GitHub Copilot Chat, without making live calls to the real Copilot API. When started and pointed at this server, the Copilot client intercepts the requests and returns scripted, pre-written answers instead of generating real AI responses. The core use case is deterministic demos: you write rules in a JSON file pairing keywords with responses. When a message contains those keywords, the mock server replies with your scripted text, streamed token by token just as Copilot would. This means screencasts or live presentations where Copilot always says exactly what you planned, with no risk of a surprise or slow response. The server handles both communication styles Copilot uses: WebSockets and HTTP server-sent events. It also works as an HTTPS interception proxy for command-line tools that respect the HTTPS_PROXY environment variable, so you can mock Copilot traffic beyond VS Code without changing code. Beyond simple scripted responses, the rules format supports multi-step sequences where Copilot appears to narrate, call tools like file creation, then narrate again, faking the look of an AI agent working through a task. Rules can include delays to simulate a thinking pause. Tool-call outputs are included, which VS Code renders as the AI actually executing file edits. A learning mode lets you record real Copilot responses while using the product normally, building up a library of actual responses to replay later. A list command prints all loaded rules before a demo so you can verify everything is correct. An option to forward unmatched prompts to the real Copilot API is also available. Setup is quick: run npx copilot-mock-server with no install required, write a rule file, and tell VS Code to proxy Copilot traffic to localhost. A helper command injects the needed VS Code settings automatically.