Agents
Choose which coding agents Proliferate can run for your team.
Proliferate runs Claude Code, Codex, Cursor, OpenCode, and Grok in the same workspaces, with the same repo context, terminals, and review surface.
There is nothing to install. Proliferate keeps its own copy of each agent pinned
to an exact version and repairs it every time the runtime starts. An agent you
installed yourself on PATH is left alone; the managed copy wins.
Supported agents
Each agent gets a page under Settings → Agents where you pick its authentication route and which models it may use.
Models
The model picker shows what the selected agent can reach on its current authentication route. Native sign-in, your own API key, and the gateway each expose a different set. There is no version to pin.
| Agent | Models |
|---|---|
| Claude Code | Anthropic's current model lineup. |
| Codex | OpenAI's current model lineup. |
| Grok | xAI's current model lineup. |
| Cursor | Cursor's own current model lineup, managed through its own account. |
| OpenCode | Whichever provider you connect, from Anthropic and OpenAI to a hundred-plus other options. |
To see exactly what is reachable right now, open Settings → Agents → [agent] → All Models.
Authentication
Model access comes from the agent's native sign-in, an API key you bring, or the Proliferate gateway. Cloud workspaces get their own authentication surface when they ship.
Admin controls
Next: set up a workspace to run an agent in.