Getting Started
From zero to your first AI agent session.
No experience required.
The Easy Way
Already using VS Code? Search, install, done.
Install the VS Code Extension
Open VS Code and search "Cohort" in the Extensions panel (Ctrl+Shift+X). Click Install.
Then open the panel:
The extension auto-installs the Cohort Python package if it's missing. The server starts automatically when you open the panel -- no setup wizard, no terminal commands.
Requires Python 3.11+ and VS Code 1.85+. Also available as a .vsix download from GitHub.
Manual Setup
Seven steps from a blank machine to a running AI agent team.
Install Python
Download Python 3.11 or newer from python.org. Run the installer and check the box that says "Add Python to PATH" -- this is the most common mistake beginners make.
After the install finishes, open a terminal and verify it worked:
On macOS or Linux, you may need to use python3 instead of python.
Install VS Code (optional but recommended)
Download VS Code from code.visualstudio.com. This gives you the full Cohort experience: channel messaging, agent panels, integrated terminals, and inline approvals.
You can skip this step and use the CLI + web dashboard instead. VS Code just makes everything more convenient.
Install Cohort
Open a terminal (Command Prompt, PowerShell, or the VS Code terminal) and run:
This installs the Cohort server, MCP tools, web dashboard, and 23 bundled AI agents. The install takes about a minute.
Windows users: The quotes around "cohort[server]" are required in PowerShell. Without them, the extras won't install and the server won't start.
Run the Setup Wizard
Run the one-time setup wizard in your terminal:
The wizard detects your hardware (CPU, GPU, available memory), configures the server, and optionally sets up Ollama for local LLM inference. Follow the prompts -- it takes about two minutes.
Start the Server
Launch the Cohort server:
The web dashboard opens automatically at http://127.0.0.1:5100. Leave this terminal window open -- it's running your server.
Install the VS Code Extension (if you installed VS Code)
Download cohort-vscode-0.4.12.vsix from the GitHub release page.
Then install it in VS Code:
Start Your First Session
Create a channel, pick your agents, and start a discussion. The orchestrator automatically picks who speaks next based on relevance scoring -- you don't have to manage turn-taking. Contribution scoring, loop prevention, and executive briefings all work out of the box.
What's Next
You're up and running. Here's where to go from here.
Full Documentation
API reference, configuration options, and advanced features.
Meet the Agents
23 pre-configured agents with specialized roles and expertise.
Channel System
How channels, messaging, and multi-agent conversations work.
VS Code Extension
Panels, permissions, session management, and keyboard shortcuts.
Troubleshooting
Common issues and how to fix them.
"python not found"
Python wasn't added to your system PATH during installation. The fix: re-run the Python installer, choose "Modify", and make sure "Add Python to PATH" is checked. Then restart your terminal.
"pip not found"
Some systems don't add pip directly to PATH. Use the long form instead:
PowerShell ignores [server] extras
PowerShell treats square brackets as wildcards. Wrap the package name in quotes: pip install "cohort[server]". Without quotes, pip installs the base package and the server dependencies are silently skipped.
"cohort" is not recognized
Python's Scripts folder isn't on your PATH. Use the module form instead:
Server won't start
Port 5100 is probably already in use by another process. Either close the other process or run Cohort on a different port. Check with netstat -ano | findstr 5100 on Windows or lsof -i :5100 on macOS/Linux.
No GPU detected
Cohort works fine on CPU -- it's just slower for local inference. If you have an NVIDIA GPU and it's not being detected, make sure your GPU drivers are up to date. You can also install Ollama manually and Cohort will detect it on next startup.
Need Help?
If you're stuck, we're here.