^AI (fai) is an open-source CLI tool that captures and shares persistent context across AI coding tools and sessions, enabling compounding intelligence without workflow changes.

---

🚀 Overview

^AI, also known as fai, is a portable, git-backed context layer for AI-assisted development. It automatically stores session notes, project knowledge, and personal preferences in plain Markdown files within your project directory, injecting this context into future interactions with any compatible AI tool.

This project solves the problem of AI silos, where each tool like Claude, Cursor, or Copilot forgets everything when you switch sessions or tools, forcing constant re-explanation of preferences, decisions, and project details. It's built for developers and teams using multiple AI coding assistants who want seamless, persistent memory that compounds over time.

Fathym built it to democratize AI context ownership, much like Git did for code—giving users full control, portability, and zero lock-in in a $29.3B market gap.

---

✨ Key Features

• Persistent shared context: Works across tools like Claude Code, Cursor, Cline, Aider, Continue.dev, GitHub Copilot, and more—no workflow changes needed.
• Automatic capture: Grabs decisions, patterns, problems solved from sessions into three vaults: Session (recent), Project (build details), Personal (your style).
• Portable storage: Plain Markdown files in your repo, git-backed, owned by you.
• Compounding intelligence: AI remembers your PostgreSQL preference or stack choices forever.
• Local-first: Runs with Ollama or OpenAI-compatible endpoints, no API keys needed for local models; MCP server for integration.

---

🛠 Tech Stack

• CLI: Deno (install via deno run -A jsr:@fathym/fai/install)
• Storage: Git-backed Markdown files (~/.fai/ vaults)
• Protocol: Model Context Protocol (MCP) server
• Integrations: Ollama, LM Studio, OpenRouter, Groq, OpenAI-compatible
• OS: Windows, macOS, Linux
• No external services/telemetry

---

👥 Who Is This For?

This project is useful for:

• Developers who want to ___stop re-explaining project details and preferences to AI tools
• Teams looking for ___shared, compounding AI intelligence across coding sessions
• Indie makers building ___with AI assistants like Cursor or Claude without lock-in

---

📸 Screenshots / Demo

• Live demo: https://www.fathym.com
• GitHub: https://github.com/fathym-deno (org; source via JSR)
• Docs: https://www.fathym.com/docs
• Install: deno run -A jsr:@fathym/fai/install

---

🔥 Project Status

• 🟢 Actively building

Last updated: 2026-03-23

---

❓ FAQ

Is this project free?

Yes — Open Source (MIT license), free to start, no API keys for local use.

Is the source code available?

Yes — https://jsr.io/@fathym/fai (MIT-licensed Deno package; GitHub repo at https://github.com/fathym-deno/power-ai linked from JSR)

Can I contribute?

Yes, check the JSR package, GitHub org https://github.com/fathym-deno for issues/PRs, or provide feedback via Twitter @FathymIt or docs.

---

📬 Contact & Links

• Website: https://www.fathym.com
• GitHub: https://github.com/fathym-deno
• Twitter/X: https://x.com/FathymIt
• Email: Not listed (contact via site/Twitter)

---

Built with ❤️ and kept warm on WarmIndex.