● Open source · MIT License

Your AI context
is portable now

Define who you are once. Export to Claude, ChatGPT, Cursor, and every AI tool. Stop re-explaining yourself.

$ pip install aura-ctx click to copy

Your context is scattered across every AI tool

You use Claude for analysis, ChatGPT for writing, Cursor for coding. Each one makes you start from scratch.

Your preferences, your stack, your style — locked inside each platform's proprietary memory, invisible to the others.

MCP solved tool interoperability.
aura solves context interoperability.

Claude
knows your code style
ChatGPT
knows your writing tone
Cursor
knows your framework
Gemini
knows nothing
None of them talk to each other.

Three commands. Every AI.

1

Define your context

Create scoped context packs — portable YAML files that describe who you are in each domain.

aura init aura create dev -t developer aura edit dev
2

Export anywhere

One command generates the right format for each tool. System prompts, .cursorrules, Claude memory, ChatGPT instructions.

aura export dev -f cursorrules aura export dev -f claude-memory aura export dev -f chatgpt-instructions
3

Or serve via MCP

Start the built-in MCP server and every compatible tool reads your context automatically. Zero config.

aura serve --mcp # Claude, Cursor, ChatGPT Desktop # all connect automatically

Your identity, structured

Context packs are human-readable YAML. Version them, share them, edit them.

~/.aura/packs/developer.yaml
name: developer scope: development facts: - key: languages.primary value: [Python, TypeScript, Rust] confidence: high - key: frameworks [Next.js 15, FastAPI, Supabase] - key: style.comments "Minimal — only for non-obvious logic" rules: - instruction: "Always use TypeScript strict mode" priority: 8 - instruction: "Prefer functional patterns over OOP" priority: 5

Export to everything

◉
System Prompt
--format system-prompt
⌘
Cursor IDE
--format cursorrules
✦
Claude Memory
--format claude-memory
◎
ChatGPT
--format chatgpt-instructions

The real power:
built-in MCP server

MCP (Model Context Protocol) is the open standard for connecting AI tools to data. aura includes a full MCP server out of the box.

Start it once. Every MCP-compatible client — Claude Desktop, Cursor, ChatGPT Desktop, VS Code, Windsurf — discovers your context automatically.

Your context packs become live resources that AI tools can query, search, and compose. No copy-paste. No re-explaining.

claude_desktop_config.json
{
  "mcpServers": {
    "aura": {
      "url": "http://localhost:3847/mcp"
    }
  }
}
terminal
$ aura serve --mcp

✦ aura MCP server running
  Serving 3 context packs
  Claude, Cursor connected

Own your context

Local-first. Open source. Your identity stays on your machine. Export it everywhere.

★ Star on GitHub Read the docs →