Get Cohort
One installer. Five minutes to your first AI agent session.
Everything runs on your machine. Your data never leaves.
pip install
For developers who already have Python. Lightweight core with only two dependencies. Full control over your setup.
- [OK] Same software, same features
- [OK] Python 3.11+ required
- [OK] Works on Windows, macOS, Linux
- [OK] CLI + MCP server
- [OK] Web dashboard included
Windows
Download, install, launch. The setup wizard handles everything -- GPU detection, model download, configuration. No terminal required.
- [OK] Embedded Python (no install needed)
- [OK] Setup wizard detects your GPU
- [OK] Downloads best model for your hardware
- [OK] System tray icon (runs in background)
- [OK] Web dashboard in your browser
- [OK] Start Menu + Desktop shortcuts
- [OK] Optional auto-start with Windows
Docker
For teams and CI/CD pipelines. Pre-built container with everything included.
- [OK] Same software, same features
- [OK] Dockerfile + docker-compose included
- [OK] Works on any Docker host
- [OK] GPU passthrough supported
What Happens After You Install
The Windows installer gets you to a working AI agent session in three steps. No terminal, no configuration files, no manual downloads.
Launch Cohort
Click the desktop shortcut or find Cohort in the Start Menu. A system tray icon appears and your browser opens to the dashboard.
Setup Wizard Runs Automatically
On first launch, the wizard detects your GPU, installs the AI engine (Ollama), and downloads the best model for your hardware. You'll see progress bars for everything -- the model is 5-8 GB depending on your GPU.
Start Your First Session
The dashboard is ready. Create a channel, pick your agents, and start a discussion. Contribution scoring, loop prevention, and executive briefings all work out of the box. After the first setup, Cohort runs entirely offline.
System Requirements
Cohort runs AI models locally on your hardware. Here's what you need.
Minimum
- OS
Windows 10 (64-bit) - RAM
16 GB - Disk
15 GB free - GPU
CPU-only mode available
Recommended
- OS
Windows 11 - RAM
32 GB - Disk
25 GB free (SSD) - GPU
NVIDIA with 8+ GB VRAM
An NVIDIA GPU dramatically speeds up AI inference (10-30x faster than CPU). The setup wizard automatically detects your GPU and selects the right model. Without a GPU, Cohort still works -- it just uses CPU inference, which is slower but fully functional.
What's in the Box
The same open-source software, packaged for convenience. Nothing is hidden, nothing is locked.
Ready to run AI agents locally?
Your agents. Your hardware. Your data.
Apache 2.0 Licensed · The installer and the pip package ship identical software · VirusTotal Verified