# The verbs

Run `ade` with nothing after it and it prints its home screen: a short status line, then every verb with a few words
on it. This is that screen, one row per verb.

| Group | Verb | What it does |
| --- | --- | --- |
| Start | `ade init` | start a project here, or install into this repo |
| Start | `ade link` | make a plain `ade` work in new shells |
| Start | `ade config [word]` | find any ade.toml setting, its value and what it does |
| Start | `ade parts` | list the kit's parts; `parts set x=off` turns one off |
| Start | `ade setup` | a panel to switch parts on and off, with the diff first |
| Work | `ade card add <title>` | a change to make, with its paths and done line |
| Work | `ade loop <card>` | lane, build, review, test, land |
| Work | `ade conduct` | loop every open card, waking a model only when needed |
| Work | `ade land <card>` | merge a reviewed, tested lane |
| Work | `ade push` | send trunk and records; fast-forward only |
| Look | `ade cockpit` | every open card and its next command |
| Look | `ade status` | the one line above |
| Look | `ade watch <card>` | follow a lane live |
| Look | `ade telemetry` | tokens, cost and week meters |
| Look | `ade budget` | lines, hooks and refusals against their limits |
| Decide | `ade decide` | rulings, and questions with a default and a deadline |
| Decide | `ade digest` | what landed, parked or defaulted today |
| Tools | `ade handover --role <r>` | an agent's starting context |
| Tools | `ade q` | store long output, then page or query it |
| Tools | `ade session` | run a verb detached, list or stop it |
| Tools | `ade wait` | wait for a condition, with a bound |
| Tools | `ade graph` | regenerate the code graph |
| Tools | `ade rehearse` | full loops on throwaway repos, spending nothing |
| Tools | `ade uninstall` | remove what init wrote, keep your records |

[Commands](https://atlas-ade.dev/docs/commands/) has the full help text for each one.

## Long output stays out of context

A test run or a build log can fill a model's context window with lines nobody needs. `ade q` keeps that out.

- Output that goes over the caps in `[q]` of `ade.toml` (`cap_lines` and `cap_bytes`) is kept whole in `.ade/q.sqlite`.
- What the agent sees is the head, the tail, the lines that look like errors, and a handle such as `q17`.
- `ade q show q17` pages through the stored output, `ade q grep q17 PATTERN` searches it, and `ade q list` lists what is stored.
- `ade q run -- CMD` runs any command and stores its output the same way. Output within the caps prints whole.
- Stored output expires after `keep_days`, 14 by default.