Now running on Apple Silicon

Your AI.
Your machine.
Your data.

Cashmere is a sovereign AI agent that runs 24/7 on your own machine. Free and open source. Zero cloud costs. Total privacy. It researches, plans, and builds while you sleep.

cashmere

$ uv run cashmere

Cashmere v0.1.0 starting...

├─ Ollama connected (gemma4:26b)

├─ Memory loaded (2,847 entries)

├─ Knowledge graph ready (12,403 nodes)

├─ Daemon scheduler active

└─ Server running on :8420

Background: Researching market trends for Q2 report...

Background: Monitoring 3 competitor feeds...

Background: Drafting weekly summary from browsing data...

Built different

Not another chatbot wrapper. Cashmere is a full cognitive system that operates autonomously on hardware you own.

Zero Cloud Costs

Run state-of-the-art models locally via Ollama. No API keys, no token meters, no surprise bills. Just electricity.

Always On

A daemon that researches, monitors, and plans 24/7. Wake up to completed tasks and fresh insights every morning.

Deep Personalization

Learns from your conversations, browsing history, and calendar. Builds a knowledge graph that grows with you.

Total Privacy

Everything — your data, your models, your conversations — stays on your hardware. No telemetry. No cloud sync.

Self-Improving

Reviews its own outputs, identifies failure patterns, and evolves its strategies. Gets better the longer it runs.

Open Source

Full source code available. Fork it, extend it, audit it. No vendor lock-in, no black boxes.

How it works

From unboxing to autonomous agent in under an hour.

01

Clone & install

Clone the repo, install Ollama, and pull the models. Five commands, ten minutes. Runs on any Mac with Apple Silicon.

02

Configure & connect

Point Cashmere at your calendar, browsing history, and data sources. Set your priorities and goals.

03

Let it run

Cashmere's daemon starts working immediately — researching, monitoring, planning. It gets smarter every day.

Free. Forever.

No tiers. No accounts. No tokens to meter. Cashmere is open source under MIT — clone it, run it, fork it, ship it.

Full source code (MIT)
Every feature, no gates
Local models via Ollama
No telemetry, no phone-home
Community support on GitHub
Run on your own hardware

We don't sell hardware and we don't sell subscriptions. You bring the machine, we bring the code.

Install in 2 minutes

macOS with Apple Silicon, 16GB+ unified memory, and Ollama installed.

terminal

$ git clone https://github.com/cashmere-ai/cashmere

$ cd cashmere

$ uv sync

$ ollama pull gemma4:26b

$ uv run cashmere

◆ Server running on :8420