You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 5b47919
Browse filesBrowse the repository at this point in the historyBrowse files
Copy file name to clipboardExpand all lines: README.md
+12-15Lines changed: 12 additions & 15 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,9 +1,10 @@
1
1
<!-- --8<-- [start:overview] -->
2
2
# agent6
3
3
4
-
- Runs model commands in a [Linux sandbox](#security)using your installed toolchains, with configurable filesystem and network access.
4
+
- Runs the model's commands in a [Linux sandbox](#security) with the file and network access you set.
5
5
- Saves conversations, tool activity, and Git checkpoints so you can inspect, resume, or fork a session from an earlier turn.
6
-
- Shows the task graph, usage, and context compaction, with pinned instructions and repository memory to carry information across context windows and runs.
6
+
- Shows the task graph, token usage and compaction as they happen.
- Verifies work with a configured or inferred command, checking the starting tree by default, and reports whether the result passed, failed, or remains unverified.
8
9
- Runs [parallel attempts][parallel-runs] in isolated clones with `--parallel N` or a list of models, then ranks the results for you to choose which to merge.
9
10
- Defines long-running workflows in editable [TOML state machines][state-machines], combining agent runs, tool calls, waits, and branches with offline tests, crash recovery, and replay from recorded results.
@@ -45,7 +46,7 @@ For subscriptions, replace `agent6 connect` with `agent6 connect chatgpt` or `ag
45
46
46
47
## [Usage][usage]
47
48
48
-
**Runs edit your current checkout by default.**
49
+
A run edits your current checkout.
49
50
Checkpoint commits go on a separate `agent6/<id>` branch; checkpointing leaves your current branch's history and index in place.
50
51
For a separate checkout, use `agent6 run --worktree "your task"`, which starts from committed HEAD without copying uncommitted changes or untracked dependencies.
51
52
@@ -70,14 +71,14 @@ Other commands:
70
71
71
72
| Command | Purpose |
72
73
| --- | --- |
73
-
|`agent6 plan "task"`|Plan a change; `agent6 run --from <plan-id>` carries it out.|
74
-
|`agent6 ask "question"`|Ask about the repository.|
75
-
|`agent6 review`|Review a diff.|
76
-
|`agent6 attach <id>`|Follow a live session.|
77
-
|`agent6 steer <id> "instruction"`|Direct the next steps.|
78
-
|`agent6 stop <id> --after-step`|Stop after the current step.|
79
-
|`agent6 resume <id>`|Continue an interrupted run.|
80
-
|`agent6 fork <id> --at-turn 7`|Fork a coding run from an earlier turn into a separate worktree.|
74
+
|`agent6 plan "task"`|plan a change; `agent6 run --from <plan-id>` carries it out |
75
+
|`agent6 ask "question"`|ask about the repository |
76
+
|`agent6 review`|review a diff |
77
+
|`agent6 attach <id>`|follow a live session |
78
+
|`agent6 steer <id> "instruction"`|direct the next steps |
79
+
|`agent6 stop <id> --after-step`|stop after the current step |
80
+
|`agent6 resume <id>`|continue an interrupted run |
81
+
|`agent6 fork <id> --at-turn 7`|fork a coding run from an earlier turn into a separate worktree |
81
82
82
83
To give a finished run more work, use `agent6 resume <id> --steer "next task"`.
83
84
@@ -104,10 +105,6 @@ The defaults are `sandbox.network = "auto"`, `sandbox.run_commands = "ask"` and
104
105
`agent6 check sandbox` shows the effective file and command boundaries, network and approval settings, and live probe results.
105
106
106
107
## Development
107
-
108
-
agent6 is [Apache-2.0 licensed](https://github.com/agent6-dev/agent6/blob/master/LICENSE) and under active development before 1.0; commands, configuration, and stored formats can change.
109
-
[Development conventions](https://github.com/agent6-dev/agent6/blob/master/AGENTS.md) cover contributions, and the [benchmark and evaluation suites](https://github.com/agent6-dev/agent6/tree/master/bench) compare models and harness settings with task graders, repeated runs, and usage measurements.
110
-
111
108
[Report bugs or confusing behavior](https://github.com/agent6-dev/agent6/issues), or [report vulnerabilities privately](https://github.com/agent6-dev/agent6/security/advisories/new).
0 commit comments