Chat Copilot

Provides a real-time streaming chat sidebar within VS Code that connects to OpenAI-compatible APIs (OpenAI, Anthropic, Google, Ollama, Azure OpenAI, DeepSeek) via configurable API endpoints and authentication tokens. Implements server-sent events (SSE) streaming to display token-by-token responses, with mid-stream interruption capability and automatic handling of truncated responses. The extension abstracts provider differences through a unified configuration layer supporting custom model names and base URL overrides.

Enables users to reference multiple files and images within a single chat conversation using @file syntax, allowing the AI to generate or modify code with awareness of existing codebase context. The extension passes selected file contents and image data as part of the chat prompt to the LLM, enabling multi-file refactoring, cross-file bug fixes, and documentation generation. Image support allows users to include screenshots, diagrams, or design mockups as context for code generation.

Manages API keys for multiple LLM providers (OpenAI, Anthropic, Google, Azure OpenAI, DeepSeek, etc.) with secure storage in VS Code's credential store. Users configure one API key per provider in extension settings, and the extension routes requests to the appropriate provider based on selected model. Credentials are encrypted and stored locally, never transmitted to third parties.

Explicitly disables all telemetry and usage data collection, ensuring user interactions, prompts, and code are never transmitted to extension maintainers or third parties beyond the selected LLM provider. This is a design choice differentiating Chat Copilot from many commercial AI tools that collect usage analytics. Users have full transparency that only LLM provider APIs receive conversation data.

Provides a Prompt Manager feature allowing users to create, save, and reuse prompt templates with #hashtag-based lookup syntax. Templates can include placeholders and are searchable within the chat interface, enabling teams to standardize AI interactions for common tasks (code review, testing, documentation). The system stores prompts locally in VS Code settings, making them available across all projects and shareable via settings sync.

Allows users to generate new files or modify existing code directly from AI responses with single-click or keyboard-shortcut actions. The extension detects code blocks in AI responses and provides inline buttons to create files, apply patches, or insert code at cursor position. This eliminates manual copy-paste workflows and integrates code generation directly into the chat-to-editor pipeline.

Enables users to export chat conversations to Markdown format for documentation, knowledge base creation, or audit trails. Conversations can be edited and resent within the chat interface, allowing users to refine prompts and regenerate responses. The extension maintains conversation history within the current session but does not persist conversations across VS Code restarts without manual export.

Supports Model Context Protocol (MCP) integration (v4.7.0+) enabling users to extend the AI's capabilities with custom tools and integrations. MCP allows the AI to call external functions, access databases, or interact with third-party services through a standardized protocol. The extension acts as an MCP client, translating tool calls from the LLM into actual function executions and returning results back to the conversation.

Supports advanced reasoning models (OpenAI o1, o3-mini, DeepSeek R1) that perform extended chain-of-thought reasoning before generating responses. These models are optimized for complex problem-solving, mathematical reasoning, and code debugging by spending more compute on reasoning steps. The extension transparently routes prompts to reasoning models and handles their longer response times and different token accounting.

Provides DeepClaude hybrid mode (v4.6.7+) combining DeepSeek R1's reasoning capabilities with Claude's code generation strengths. This mode routes complex reasoning to DeepSeek R1, then uses Claude to generate final code based on reasoning output, optimizing for both problem-solving depth and code quality. The extension manages the multi-model pipeline transparently within a single conversation.

Integrates GitHub Copilot as a selectable LLM provider (v4.6.9+) within the Chat Copilot interface, allowing users to route conversations through GitHub Copilot's models instead of OpenAI or other providers. This enables teams already invested in GitHub Copilot to use Chat Copilot as a unified chat interface while maintaining their existing Copilot subscription and authentication.

Supports local model execution via Ollama integration, allowing users to run open-source models (Llama, Qwen, DeepSeek R1, etc.) on their own hardware without cloud API costs. Users configure a custom model name and point to a local Ollama instance (default localhost:11434), enabling fully offline operation and complete data privacy. The extension treats Ollama as an OpenAI-compatible API endpoint, abstracting the local execution details.