Self-hosted AI agents
you actually own.
Sandboxed, long-lived AI agents on your own infrastructure. Provisioned, scaled, paused, and watched from a single self-hosted console.
Sandboxed, long-lived AI agents on your own infrastructure. Provisioned, scaled, paused, and watched from a single self-hosted console.
One API. Swappable brains.
Multi-tenant workspaces
Agents, keys & data isolated per team or client
owner · admin · member Plain Docker · REST + SSE API · Self-hosted · Your keys, encrypted · Provision in under 5 seconds
Each agent keeps a writable /workspace volume across restarts, pauses, and weeks of idle. Files, memory, and history persist.
Every agent runs in its own hardened container: read-only root, dropped capabilities, egress filtering. An unhinged agent can't touch the host or its neighbours.
One identical API, swappable execution engines. Vanilla, Opencode, Codex, or Claude Code. Pick the right brain per agent and switch anytime.
Single-VM deploy, your keys, your data. No vendor lock-in, no per-seat tax, nothing leaving your stack. One command: make prod.
Provision from a template in seconds. Idle agents auto-pause to save resources and auto-wake on the next task. The fleet breathes on its own.
Learn more about persistent agents →A built-in tool-use loop you fully control: pick the tools, write the system prompt, tune budgets.
An embedded coding harness that manages its own tools and context.
OpenAI's Codex agent, with skill support.
Anthropic's agent via claude -p. Use an API key or sign in with your own Claude plan, then assign it to any agent in the fleet.
Read-only root filesystem, dropped Linux capabilities, egress filtering (private ranges & cloud metadata blocked), CPU/memory limits.
Bring any provider's API key, or connect your Claude and ChatGPT subscriptions. Kept on your VM, revocable in a click. No per-seat SaaS tax.
Automatic snapshots after every task, non-destructive rollback to any point, and optional Git backup that auto-pushes on completion.
Everything the web console does is a first-class REST call with a Server-Sent-Events stream. Create named, revocable API keys and drive your whole fleet from your backend.
Read the API docs → Learn more about self-hosting →Self-hosted infrastructure for a fleet of long-lived AI agents. Each agent runs in its own sandboxed container with a persistent workspace, internet access, and a pluggable driver.
Yes. Single-VM deploy on plain Docker with your own keys and data. No vendor lock-in, no per-seat SaaS tax, nothing leaves your stack.
One identical API across swappable brains: Vanilla, Opencode, Codex, and Claude Code. The driver is part of an agent's config and can be swapped anytime.
Every agent runs in a hardened container: read-only root filesystem, dropped capabilities, network egress filtering, and CPU/memory limits. LLM keys are stored server-side, encrypted, and never sent to the browser.
Yes. Every task streams over Server-Sent Events (assistant messages, tool calls, results, file changes, a live token meter), and reconnects resume where they left off.
Self-host it. Clone the repo and run make dev to try it locally, or make prod to deploy on your own VM. Admins create accounts on your install, and there is no public sign-up.
Put a fleet of sandboxed agents to work on your own infrastructure, provisioned in seconds and watched live from one console.
Admin-provisioned · Self-host in one command · Your data never leaves your VM