# The OrbitMap work model

OrbitMap gives a person and an AI agent a shared map of the work. You can talk to
the agent in plain language, while the app keeps the outcome, decisions and open
work in a place that survives the chat. The names below are useful when you need
to find something or decide where a new piece of work belongs; they are not a
form you have to fill in all at once.

## Start with the level you need

**Workspace** is the team's shared home. It brings together the repositories,
current focus, agreed outcomes and project memory. Use it when work crosses
repositories or when you want to see the whole team rather than one codebase.

**Area** is the part of the workspace where implementation happens, usually one
repository. An intent can span several areas, but each task belongs to the area
whose code or files it changes. You normally do not choose an area in every
conversation: after `init`, OrbitMap resolves it from the folder where your
agent is working. In the app, areas make it clear which product or repository a
piece of work touches.

**Mission** is the team's current direction for a limited period. It answers
“what are we trying to move forward now?” Use one to keep several related
intents together, for example “make the first week successful for new teams.”
The mission is a lens over the workspace, not a second task list.

**Intent** is one outcome you are willing to stand behind. It might be “a new
team can connect its repositories and reach its first useful view.” An intent
holds the promise, the design agreed with you, the plan and the record of what
happened. It is the right level when the work needs discussion or a decision
that should not be rediscovered in another chat.

**Task** is a bounded piece of an approved intent that an agent can build and
check. Tasks are created from the plan and placed in areas. You steer the
intent; the agent uses tasks to carry it through. You do not need to turn every
small thought into a task—use an intent when there is an outcome to agree on,
and let the plan provide the smaller pieces.

## What this looks like in the app

The workspace view gives you the broad picture: the active mission, intents and
the areas connected to the work. The mission view groups the outcomes currently
important to the team. An intent's planning view is where you can follow its
outcomes, design, plan and work log; after the build, it becomes the place to
review what was delivered.

The same records are available to your agent through the OrbitMap CLI. Run the
agent from the workspace or an area folder after [setting up the connection](/docs/setup-and-troubleshooting).
Commands such as `orbitmap context` show what that
folder resolves to. You can ask naturally:

```text
Show me the active mission and the intents connected to it.
```

```text
Find the intent about onboarding and tell me what is waiting for review.
```

The agent reads the shared records before proposing work, and writes accepted
outcomes, decisions, tasks and useful hand-off notes back to the relevant
record. That is why another person—or another agent session—can continue
without relying on one conversation's memory.

## A small example

Imagine a team improving its trial signup. The workspace contains the web app
and billing service as two areas. The mission is “make the first week easy.”
“Let a new customer create a workspace without support” is an intent. Its plan
may produce tasks for the signup screen, the account endpoint and the welcome
email, each in the area where that work lives.

If the agent asks whether an invite should expire after 24 or 72 hours, that is
an intent-level decision: you answer once, and the choice remains visible with
the design. If it asks which helper function to edit while executing a task,
that is ordinary implementation detail. The distinction keeps OrbitMap useful
for the decisions people need to see, without making you manage every line of
the build.

## Next

- [The intent flow](/docs/intent-flow): how a conversation becomes reviewed work.
- [Working with your agent](/docs/working-with-agents): tasks, memory and hand-offs.