chat.md: The Hacker's AI Chat Interface [Experimental]
@rusiaaman
About chat.md: The Hacker's AI Chat Interface [Experimental]
An md file as a chat interface and editable history in one.
Overview
What is chat.md?
chat.md is a Visual Studio Code extension that reimagines AI interaction through plain text files. It treats conversations with AI as markdown files with a .chat.md extension, allowing editing, version control, and sharing like any other file.
How to use chat.md?
Install chat.md from the VS Code marketplace, then configure your API key(s) via Command Palette → "Add or Edit API Configuration". Create a new chat with Opt+Cmd+' (Mac) or Ctrl+k Ctrl+c (Windows/Linux). Write your query in a # %% user block and press Shift+Enter to stream the AI response. Optionally start a live markdown preview side by side.
Key features of chat.md
- File-based conversations stored as local
.chat.mdfiles - Universal model support: Anthropic, OpenAI, and any OpenAI-compatible API
- Universal MCP tool ecosystem – any LLM can use any MCP server
- Contextual file attachments using markdown syntax
- Editable, non-linear conversations – rewrite history or branch freely
- Resume incomplete AI responses at any point
- Git-friendly: track changes, collaborate, and version control conversations
Use cases of chat.md
- Non-linear editing and rewriting of AI conversation history
- Attaching project files directly for contextual AI assistance
- Creating reusable conversation templates for common tasks
- Switching between AI models mid-conversation without losing tool functionality
- Collaborating on prompts using version control and diffs
FAQ from chat.md
What models does chat.md support?
chat.md supports Anthropic Claude (Opus, Sonnet, Haiku), OpenAI GPT models, and any OpenAI-compatible endpoint (including Azure, Google Gemini, and local models via Ollama/LLM Studio).
Does chat.md support MCP servers?
Yes. chat.md is a Model Context Protocol client that supports both local (stdio) and remote (SSE) MCP servers. Any AI model can use MCP tools without vendor lock-in.
How do I create and run a conversation?
Create a file with a .chat.md extension, write a query in a # %% user block, and press Shift+Enter. The AI streams its response directly into the file.
Is chat.md free to use?
Yes, chat.md is MIT licensed and available for free from the VS Code marketplace. You are responsible for your own API keys and usage costs.
What are the current limitations of chat.md?
Only MCP tools are supported (prompts and resources are planned for the future). Caching is not yet supported in the Anthropic API. Some models (Gemini, Ollama) must be accessed via OpenAI-compatible API. The extension is considered a feature-rich proof of concept.
Basic information
Comments