Task CLI reference
Start, inspect, continue, cancel, apply, and publish proof-bearing Tasks.
Use this page to inspect every public atoi task command, its accepted flags, and one runnable example.
atoi taskStart, inspect, continue, cancel, apply, and publish proof-bearing Tasks.
| Command | What it does |
|---|---|
atoi task list | List Tasks, optionally within one Project. |
atoi task start | Start durable implementation work in a Project. |
atoi task get | Read Task state, result, changes, checks, asks, and proof. |
atoi task changes | Inspect the materialized, proof-linked Task change set. |
atoi task apply | Apply a verified change set to the matching local repository. |
atoi task draft-pr | Publish a verified change set to a proof-backed draft pull request. |
atoi task wait | Poll until a Task needs you, completes, or fails. |
atoi task cancel | Cancel a Task that is still active. |
atoi task continue | Continue completed or interrupted work with a new objective. |
#task list
List Tasks, optionally within one Project.
atoi task list [--project <project-id>] [--limit <n>]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--project | string | Select a Project by id. | ||
--limit | number | Limit the number of returned records. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task list
atoi task list --project prj_8k2v --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "list",
"project_id": "prj_8k2v",
"limit": 10
}
}
#task start
Start durable implementation work in a Project.
atoi task start --project <project-id> (--prompt <text> | --brief <text>) [--title <title>] [--model <model-id>] [--profile <profile>] [--place <place>] [--computer <computer-id>] [--confirm]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--project required | string | Select the Project that owns the Task. | ||
--prompt | string | State the Task objective. | ||
--brief | string | State the Task objective with the compatibility flag. | ||
--title | string | Set a Task title. | ||
--model | string | Select a model id. | ||
--profile | string | Select the Task safety profile. | ||
--place | string | Choose cloud, none, host, or a Computer name. | ||
--computer | string | Select an exact Computer id for host placement. | ||
--confirm | boolean | Confirm a placement choice when Atoi requires it. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task start
atoi task start --project prj_8k2v --prompt "Audit the release state and return evidence." --profile safe --place cloud --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "start",
"project_id": "prj_8k2v",
"prompt": "Audit the release state and return evidence.",
"safety_profile": "safe",
"idempotency_key": "task_release_1"
}
}
#task get
Read Task state, result, changes, checks, asks, and proof.
atoi task get <task-id> [--task <task-id>]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--task | string | Select a Task by id instead of a positional id. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task get
atoi task get tsk_3m9q --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "get",
"task_id": "tsk_3m9q"
}
}
#task changes
Inspect the materialized, proof-linked Task change set.
atoi task changes <task-id> [--task <task-id>]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--task | string | Select a Task by id instead of a positional id. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task changes
atoi task changes tsk_3m9q --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "changes",
"task_id": "tsk_3m9q"
}
}
#task apply
Apply a verified change set to the matching local repository.
atoi task apply <task-id> --confirm --expected-change-set-id <id> --expected-change-set-fingerprint <fingerprint> [--directory <path>] [--idempotency-key <key>]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--task | string | Select a Task by id instead of a positional id. | ||
--confirm required | boolean | Confirm the exact consequential action after you inspect its target. | ||
--expected-change-set-id required | string | Confirm the change set id returned by task changes. | ||
--expected-change-set-fingerprint required | string | Confirm the change set fingerprint returned by task changes. | ||
--directory | string | current directory | Apply inside this matching local repository. | |
--idempotency-key | string | Reuse a mutation identity when you safely retry the same action. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task apply
atoi task apply tsk_3m9q --confirm --expected-change-set-id chg_7p4d --expected-change-set-fingerprint 0123456789abcdef --directory . --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "apply",
"task_id": "tsk_3m9q",
"confirmation": true,
"change_set_id": "chg_7p4d",
"change_set_fingerprint": "0123456789abcdef",
"applied_revision": "6d41b2e",
"idempotency_key": "task_apply_1"
}
}
The terminal command writes the change set into the directory you name. Through MCP, your client applies the files and the call records the revision it produced.
#task draft-pr
Publish a verified change set to a proof-backed draft pull request.
atoi task draft-pr <task-id> --confirm --expected-change-set-id <id> --expected-change-set-fingerprint <fingerprint> [--head <branch>] [--title <title>] [--body <body>] [--idempotency-key <key>]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--task | string | Select a Task by id instead of a positional id. | ||
--confirm required | boolean | Confirm the exact consequential action after you inspect its target. | ||
--expected-change-set-id required | string | Confirm the change set id returned by task changes. | ||
--expected-change-set-fingerprint required | string | Confirm the change set fingerprint returned by task changes. | ||
--head | string | Choose the draft pull request branch. | ||
--title | string | Set the draft pull request title. | ||
--body | string | Set the draft pull request body. | ||
--idempotency-key | string | Reuse a mutation identity when you safely retry the same action. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task draft-pr
atoi task draft-pr tsk_3m9q --confirm --expected-change-set-id chg_7p4d --expected-change-set-fingerprint 0123456789abcdef --head task/release-audit --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "draft-pr",
"task_id": "tsk_3m9q",
"confirmation": true,
"change_set_id": "chg_7p4d",
"change_set_fingerprint": "0123456789abcdef",
"head_branch": "task/release-audit",
"pull_request_title": "Audit release state",
"pull_request_body": "Adds the verified release audit.",
"idempotency_key": "task_pr_1"
}
}
#task wait
Poll until a Task needs you, completes, or fails.
atoi task wait <task-id> [--task <task-id>] [--timeout-ms <ms>] [--interval-ms <ms>]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--task | string | Select a Task by id instead of a positional id. | ||
--timeout-ms | number | 900000 | Stop polling after this many milliseconds. | |
--interval-ms | number | 2000 | Wait this many milliseconds between reads. | |
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task wait
atoi task wait tsk_3m9q --timeout-ms 300000 --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "get",
"task_id": "tsk_3m9q"
}
}
#task cancel
Cancel a Task that is still active.
atoi task cancel <task-id> [--task <task-id>]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--task | string | Select a Task by id instead of a positional id. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task cancel
atoi task cancel tsk_3m9q --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "cancel",
"task_id": "tsk_3m9q",
"confirmation": true,
"idempotency_key": "task_cancel_1"
}
}
#task continue
Continue completed or interrupted work with a new objective.
atoi task continue <task-id> (--prompt <text> | --brief <text>) [--title <title>] [--model <model-id>] [--profile <profile>] [--place <place>] [--computer <computer-id>] [--confirm]
| Flag | Type | Default | Values | What it does |
|---|---|---|---|---|
--task | string | Select a Task by id instead of a positional id. | ||
--prompt | string | State the continuation objective. | ||
--brief | string | State the continuation objective with the compatibility flag. | ||
--title | string | Replace the Task title. | ||
--model | string | Select a model id. | ||
--profile | string | Select the Task safety profile. | ||
--place | string | Choose cloud, none, host, or a Computer name. | ||
--computer | string | Select an exact Computer id for host placement. | ||
--confirm | boolean | Confirm a placement choice when Atoi requires it. | ||
--json | boolean | false | Emit the canonical JSON envelope. | |
--ndjson | boolean | false | Emit one canonical JSON envelope per line. | |
--format | string | text | text, json | Choose human text or the JSON envelope. |
--help -h | boolean | Show command help. |
#Run task continue
atoi task continue tsk_3m9q --prompt "Apply the review findings and rerun checks." --place cloud --json
{
"tool": "atoi_tasks",
"arguments": {
"action": "continue",
"task_id": "tsk_3m9q",
"prompt": "Apply the review findings and rerun checks.",
"idempotency_key": "task_continue_1"
}
}
Read Run work from the terminal for the rest of this group, or the CLI reference index for every command.
Was this useful?