Get started
Two commands and a browser tab.
Install Prokop, initialize your workspace, connect a model, and start your first session.
1. Install
Pick your platform. The copy button grabs the active command.
The installer downloads the latest binary to ~/.prokopai/bin/prokop (on Windows: %USERPROFILE%\.prokopai\bin\prokop.exe) and adds it to your PATH.
2. Set up and open
prokop init does the whole first run: creates the standard directories under ~/.prokopai/, sets up the SQLite database, installs the bundled preconfigs, starts the daemon on port 8742, and opens the client in your browser. The coding tools agents need are built into the binary. Nothing else to install.
3. Connect a model
In the client that just opened, go to Settings → Providers. Paste an API key for your provider, or connect a supported subscription. Then go to Models and choose your default provider and model. You can still pick a different model per session.
Connect a provider and choose a default model → Watch the 35-second walkthrough, with written steps for subscriptions and API keys.
- API key: OpenAI, DeepSeek, OpenRouter, MiniMax, Zhipu, Zhipu Coding Plan
- Subscription: Codex (ChatGPT)
- Gateway: OpenRouter accepts provider/model IDs supported by its API
You never need to touch a .env file for this flow. API keys can also live there alongside server settings such as the auth token or TLS described in the docs.
4. Start a session
On first launch, Prokop creates a virtual workspace in a UUID-named directory under ~/.prokopai/workspaces. You don't need to add a project to start.
- Use the virtual workspace already available.
- Pick a model and a preconfig if you want one.
- Start chatting. The agent can read, edit, search, and run commands right away.
Want to work on an existing project instead? Add your local project folder → This optional walkthrough takes 19 seconds.
Everyday commands
The daemon keeps running when you close the terminal or the browser tab. Open the same workspace again from any device on your network.
Where to go next
- How-to guides: short videos and written steps for specific tasks
- Documentation: behavior, settings, and commands
- Configuration reference: models, providers, server settings, compaction
- Workspaces and sessions: memory, skills, goals, subagents