How Statica works

Statica is a distributed platform. The web app you see is just the front of house — real work is split across three pieces. The Statica server owns the data: workspaces, issues, members, the task queue. The daemon runs on your own machine, polls for tasks, and drives the AI coding tool. The AI coding tool — Claude Code, Codex, Cursor, Copilot, and friends — is the thing that actually writes code.

That separation is the biggest difference between Statica and tools like Linear or Jira. Agents don't execute on our servers. They run on your machine, against your repository, with your keys.

The server is also a realtime hub: it pushes WebSocket updates so the rest of the team sees activity as it happens. It never executes agent tasks itself.

The daemon ships inside the Statica CLI. On start it detects which AI coding tools are installed locally, registers with the server, and begins polling every few seconds with regular heartbeats so the server knows it's healthy.

Because the toolchain stays local, your API keys, working directory, and authorized tools never leave your machine — whether you self-host or use Statica Cloud.

Take the most common case: you assign an issue to an agent.

• You click assign in the web UI. The browser sends a request to the server.
• The server marks the agent as assignee and enqueues an execution task with status queued.
• The daemon on your machine picks it up on its next poll (usually within a few seconds). Status becomes dispatched.
• The daemon creates an isolated working directory and invokes the right AI coding tool. Status becomes running.
• The AI writes code, runs tests, and streams comments back to the server.
• When execution ends, the daemon reports success or failure. Status becomes completed or failed, and the web UI updates in realtime.

For the deeper mechanics, see Daemon and runtimes and Tasks.

Assigning an issue is the most common trigger, but it isn't the only one. Statica has four, one per collaboration style:

• Assign an issue. Hand a ticket to an agent the same way you'd hand it to a teammate.
• @-mention in a comment. Don't change the assignee or status — just ask the agent to take a look.
• Direct chat. Standalone conversation, not tied to an issue. Ask questions, ask it to draft an issue, brainstorm.
• Autopilots. Standing instructions on a schedule — a weekly digest, a Monday standup summary, a nightly triage pass.

A runtime is the pairing of one daemon with one AI coding tool. A single machine with both Claude Code and Codex installed, joined to two workspaces, registers four independent runtimes.

Local daemon runtimes are the supported model today. Hosted cloud runtimes — where you don't need a machine of your own running — are on the roadmap.