Friday

Your local AI assistant that actually does things.

Friday is an open-source, autonomous AI assistant that lives on your machine.
It reads your files, runs your commands, searches your codebase, listens to your voice — and does it all through a native desktop app or a terminal.

Download · Features · Skills · Setup for Devs · Architecture · Roadmap

📥 Download

No Python. No Node.js. No terminal. Just install and go.

1. Head to the Releases page.
2. Grab the latest .msi or .exe installer.
3. Run it. Friday appears in your system tray — ready to work.

The desktop app is built with Tauri and ships with a bundled Python sidecar, so everything runs out of the box.

---

✨ What It Can Do

Capability	How it works
Autonomous task execution	Give Friday a goal. It breaks it down, runs shell commands, edits files, handles errors — all on its own.
Interactive Permissions	Total control over Friday's actions. Approve every shell command or file operation via an aesthetic UI, set custom whitelists, or use Turbo mode.
Multi-Agent Swarms	Use /swarms to delegate complex tasks to specialized, concurrently running sub-agents (e.g. 'Coder', 'Researcher').
Custom Native Skills	Drop a .md file into ~/.friday/skills — it instantly becomes a slash command with zero coding. Learn more ↓
Background Tasks	Use /goal for infinite autonomous loops or /schedule for cron-based background execution.
Smart Action Button	The send button transforms into Stop while Friday is busy, and into Queue when you start typing mid-task.
Agent Dashboard	See what the agent is thinking, what tools it's calling, and its execution plan in real-time.
Rich Artifacts Viewer	Renders Mermaid diagrams, code diffs, and structured outputs natively inside the app.
Voice input	Press Ctrl+Alt+Space or say "Friday" / "Hey Friday" / "Пятница" from anywhere to talk to the assistant.
Vision	Friday takes screenshots and analyzes them — debug UI bugs, read diagrams, understand context.
Semantic code search (RAG)	Your entire workspace is indexed locally with ChromaDB. Ask questions about code in natural language.
Plugins & MCP	Extend Friday with Model Context Protocol servers — GitHub, Jira, databases, web search.
Dual memory	Conversation history + persistent workspace knowledge.
Any LLM	OpenAI, Claude via OpenRouter, or fully offline with Ollama (with Native Tool Calling). Switch models on the fly.
Fault-Tolerant	Gracefully handles LLM network drops, repairs broken JSON responses, and prevents Unicode crashes.

---

🧠 Custom Skills

The most powerful way to personalise Friday. Teach her a role, a style, a context — anything you need.

How it works:

1. Create a .md file anywhere in ~/.friday/skills/
2. Write your system instructions inside — role, tone, constraints, context
3. Friday picks it up automatically. No restart required.

Type / in the chat input to open the skills menu and see all available commands:

/reviewer   → activates code review mode
/architect  → switches Friday into system design mode
/writer     → enables technical documentation style
/your-skill → whatever you create

Skills are injected directly into the agent's system context the moment you trigger them. They stack with Friday's built-in capabilities — vision, RAG, tools — everything still works.

---

⚙️ Configuration

Everything is configured inside the app. No .env files needed.

1. Open Friday → click the ⚙ gear icon in the sidebar.
2. Pick your provider (openai / openrouter / ollama).
3. Paste your API key, choose a model.
4. Hit Save.

💡 Tip for free top-tier LLMs: Use OmniRoute — the easiest way to connect frontier models to Friday without expensive API keys. Run it locally and point Friday to its endpoint.

Changes apply instantly via hot-reload — no restart required.

🔧 Advanced: environment variables (for terminal/headless mode)

cp .env.example .env

Variable	Example	Description
FRIDAY_LLM_PROVIDER	openrouter	LLM backend
FRIDAY_LLM_API_KEY	sk-or-v1-...	Your API key
FRIDAY_LLM_MODEL	openai/gpt-4o	Model identifier
FRIDAY_LLM_BASE_URL	http://localhost:11434	Custom endpoint (Ollama)

---

🏗️ Architecture

Friday
├── src/
│   ├── api/          # FastAPI backend (sidecar engine)
│   ├── cli/          # Terminal REPL interface
│   ├── core/         # Agent loop & orchestration
│   ├── daemon/       # System tray, hotkeys, file watchers
│   ├── executor/     # Sandboxed command & file execution
│   ├── llm/          # LLM abstraction (OpenAI, Ollama, OpenRouter)
│   ├── memory/       # Conversation + workspace memory
│   ├── planner/      # Task decomposition & autonomous planning
│   ├── plugins/      # MCP plugin manager
│   ├── retrieval/    # RAG with ChromaDB + sentence-transformers
│   ├── speech/       # Voice recognition subsystem
│   ├── tools/        # Built-in tool registry
│   ├── ui/           # Tauri + React + Vite frontend
│   └── vision/       # Screenshot capture & analysis
├── tests/            # 189 tests, fully typed
└── pyproject.toml

graph LR
    A[Desktop App / Terminal] --> B[Friday Runtime]
    B --> C[Task Planner]
    C --> D[Execution Engine]
    D --> E[Tool Registry]
    E --> F[Shell]
    E --> G[Files]
    E --> H[Vision]
    E --> I[RAG Search]
    E -.-> J[MCP Plugins]
    D --> K[LLM Layer]
    K --> L[OpenRouter / OpenAI]
    K --> M[Ollama]
    B --> N[(Memory)]

Loading

Security: All command execution and file access is logged. Destructive operations can require explicit user approval.

---

💻 Usage

Desktop App

The primary way to use Friday. Runs as a native window with a glassmorphism dark-theme UI. The Python backend operates invisibly as a sidecar process — no terminal windows, no setup.

Slash Commands

Type / in the chat input to open the command menu:

Command	Action
/voice	Speak your request
/goal	Start an infinite autonomous task loop
/schedule	Run a command on a cron schedule
/swarms	Delegate task to multi-agent sub-agents
/<skill>	Trigger any custom skill from ~/.friday/skills
/clear	Reset conversation
/exit	Shut down

Voice & Background Daemon

Friday runs quietly in your system tray and listens for voice commands system-wide.

• Wake Words: Say "Friday", "Hey Friday", "Пятница", or "Эй, пятница" from anywhere.
• Global Hotkey: Press Ctrl+Alt+Space to instantly activate voice input.

Terminal REPL

For those who prefer the command line:

friday

---

🔨 Building from Source

Prerequisites

• Python 3.12+
• Node.js 18+
• Rust toolchain (for Tauri)

# Clone
git clone https://github.com/Cristofervaltz/Friday.git && cd Friday

# Python backend
python -m venv .venv
.venv\Scripts\activate        # Linux/macOS: source .venv/bin/activate
pip install -e .[gui,speech,vision,rag,dev]

# Frontend (Tauri)
cd src/ui && npm install

# Build the Python sidecar executable (required for production build)
# python scripts/build_sidecar.py

npm run tauri dev              # Development mode (uses local python)
npm run tauri build            # Production .exe installer (requires sidecar)

The desktop app bundles the Python backend into a highly optimized sidecar (friday-api.exe), meaning end-users don't need Python installed. It also features:

• In-App Settings: Configure API keys, models, and providers directly via a beautiful UI modal.
• Task Queue & Instant Send: View queued tasks and click ⚡ to immediately jump a task to the front of the queue.
• Workspace Selector: Set active context directories directly from the header dropdown.
• Hot-Reloading: Change your LLM or API keys and apply them instantly without restarting.
• Native System Tray: Runs quietly in the background and can be summoned anytime.

Running CI checks locally

black .          # Formatting
ruff check .     # Linting
mypy src tests   # Type checking
pytest           # 189 tests

---

🗺️ Roadmap

Stage	Status
Core runtime, LLM abstraction, memory	✅ Done
Task planner & autonomous execution	✅ Done
Speech recognition	✅ Done
Plugin system & MCP support	✅ Done
Vision & RAG (semantic code search)	✅ Done
Background daemon & system triggers	✅ Done
Native desktop app (Tauri)	✅ Done
In-app settings & hot-reload config	✅ Done
Fault-Tolerance & Ollama Native Tool Calling	✅ Done
Custom Skills & slash command menu	✅ Done
Global integrations & expansion	🚧 Next

See the full development roadmap for details.

---

🤝 Contributing

Contributions welcome. The project uses strict linting and 100% CI enforcement:

• Black for formatting
• Ruff for linting
• mypy (strict mode) for type safety
• pytest for tests

All checks must pass before merge. See the CI workflow.

---

_{Open source · MIT License · Made by @Cristofervaltz}