Introduction
Orbitmap is a shared place for deciding what to build with an AI coding agent, then keeping the decisions and unfinished work available to the next session. It connects your repositories to a workspace, while the agent works through agreed outcomes instead of jumping straight from a request to code.
The useful mental model
- A workspace is the team-level home for missions, intents, tasks and documents.
- An area is a repository connected to that workspace. A workspace can contain one area or several.
- A mission is a short-lived focus for the workspace.
- An intent is an outcome you agree with your agent before design and implementation begin.
- A task is a concrete piece of an intent's planned work.
These are not extra project-management forms to maintain for their own sake. They give the agent a durable answer to “what are we trying to achieve, what did we decide, and what is still open?”
What the app and agent each do
The web app is where people see the workspace, areas, missions and intent progress, and review what the agent proposes or delivers. Your coding agent is where the conversation happens. Claude Code and Codex use the Orbitmap CLI and its installed instructions to read context, ask the next useful question, make changes and leave a work log.
You can start in the app and copy the connection command, or run the CLI directly in the folder containing your code:
npx orbitmap init
The command connects the folder, installs the agent instructions and records the local mapping. It does not upload your source code. After it finishes, open Claude Code or Codex in the same folder - in a terminal, in the Claude app's Code tab, or in Codex in the ChatGPT app - and ask what to build first. The agent reads the repository locally and creates or continues Orbitmap work through the normal gates.
The first connection is intentionally simple: follow onboarding until the connection step, run init, then return to the app to see the connected agent and areas. The app presents the one-time choices when they matter.
The flow after connection
The agent turns an idea into an agreed outcome, then a design, a plan and implementation. At each gate it records enough context for another person or another session to continue. Work normally ends in review, where you decide whether the result is ready.
Read Getting started for the short path from onboarding to the first agent conversation. For connection details and recovery, see Setup and troubleshooting.