Open-source agent orchestration
Turn a prompt into a pull request. Put a workflow on a schedule. Pick up a session from your phone.
Optio runs coding agents and automations on your Kubernetes cluster or your own machines—with one place to follow the work.
jonwiggins/optio · Self-hosted · MIT licensed · Web, iOS & Android
Your agent. Your choice.
Built around the work
Optio handles the environment, triggers, and follow-through around your agents. You decide what to delegate and when to step in.
Give an agent a repository and a bounded change. It works in a git worktree, opens a PR, and can resume with review feedback while Optio follows CI and merge status.
Schedule dependency updates, prepare a daily briefing, or investigate an incident. Turn event payloads into prompt parameters and keep the logs and outcome of every run.
Give persistent agents a name, instructions, and an inbox. They work in turns, wake on messages or triggers, and can coordinate with other agents in your workspace.
Inside Optio
One feed for tasks, recurring workflows, sessions, and persistent agents. Open the conversation, follow a run, or return to the terminal.
Keep one-off tasks, recurring automations, local sessions, and persistent agents in the same feed. See what is running, what finished, and what needs you.
Choose When, Where, Who, What, and Then. Start with the trigger, choose a pod or your machine, and pick the agent and prompt. A plain-language summary tells you what will happen.
Read the conversation, use the terminal, or open a shell in the same directory. Split sessions side by side and resize the session list. Your pane arrangement stays on this device.
Create named specialists with their own instructions and inboxes. They can wake on messages, work together through the inter-agent API, and wait between turns.
Follow a run through its logs and results. Track PRs, review feedback, and costs where available, then return to the work when you need the context again.
Real Optio screens with fictional example data. Activity and usage in this tour are simulated.
Your environment, connected
Keep repeatable work in your cluster. Pair your laptop for hands-on sessions. Bring the tools your agents need.
Explore the architecture ↗Manual · Schedule · Webhook · Ticket
GitHub · GitLab · Slack · Linear · Jira · Pylon · PagerDuty · Sentry · Alertmanager · Datadog
Explore integrations ↗Web to pocket
Native iOS and Android apps keep the same workspace within reach. Check the queue, reply to an agent, open a terminal, or start new work.
Native apps built from source. Screens show fictional example work.
A closer look
From the event that started a run to the conversation that moves it forward. Open a screen to explore it at full size.
Before you start
Want the implementation details?
Find them in the documentation ↗A Kubernetes cluster, PostgreSQL, and Redis run the control plane. The local setup script deploys these with Docker Desktop. Bring credentials for the agent runtime you want to use; Optio Local can also connect your own machines for execution.
Choose a runtime for each piece of work, with the models and authentication methods that runtime supports. On a paired machine, Optio Local uses the CLI configuration and login already installed there. You can also run a shell command without an LLM.
No. Repository work can produce pull requests, while standalone work can query connected tools, generate reports, or run commands. Interactive sessions and persistent agents support other ways of working. The Work form offers the combinations that fit.
Choose the connections, tools, and skills available to pod work. Configure code review, draft PRs, and automatic resume limits. Per-work settings can tighten a repository’s PR policies. Use organization or private ownership and workspace roles to manage access.
Optio is open source under the MIT license. You operate its infrastructure and store its state. Your chosen agents still communicate with their model providers and connected services; those services and your infrastructure may have their own costs.
Start with one piece of work
Deploy locally with Docker Desktop and Kubernetes, connect an agent, and give it something useful to do.
git clone https://github.com/jonwiggins/optio.git
cd optio
./scripts/setup-local.shRequires Docker Desktop with Kubernetes, Node.js 22+, pnpm 10+, Helm 3+, and kubectl.
Full installation requirements ↗