Run mission workflows with the Praxa CLI

Mission commands call your Integration Gateway through @praxa/sdk. Configure the Gateway origin and delegated OAuth token before starting.

export PRAXA_BASE_URL="https://your-gateway.example"
export PRAXA_ACCESS_TOKEN="<short-lived delegated OAuth token>"

Choose intent or mission intake

Goal	Command	Result
Record natural-language intent for deterministic compilation	`mission submit`	Pending intent submission
Submit an explicit goal and resource budget	`mission create`	Initial mission projection

Neither admission response proves completion.

Submit intent

Generate one stable key for this logical submission and keep it for retries:

IDEMPOTENCY_KEY="intent:$(uuidgen | tr '[:upper:]' '[:lower:]')"

praxa mission submit \
  --intent "Prepare a governed release checklist" \
  --idempotency-key "$IDEMPOTENCY_KEY" \
  > intent.json

The intent is limited to 4,000 characters. If the command loses its response, rerun the exact text with the same key.

Create a budgeted mission

mission.json

{
  "goalSpec": {
    "task": "Prepare the weekly engineering review"
  },
  "resourceBudget": {
    "maximumSteps": 12,
    "maximumToolCalls": 8,
    "maximumElapsedMs": 120000,
    "maximumParallelism": 2
  }
}

IDEMPOTENCY_KEY="mission:$(uuidgen | tr '[:upper:]' '[:lower:]')"

praxa mission create \
  --input mission.json \
  --idempotency-key "$IDEMPOTENCY_KEY" \
  > mission-admission.json

Validate mission.json before the command. The CLI reads and parses the file, then the SDK and Gateway validate the contract.

Read authoritative state

--run-id must be a UUID:

RUN_ID="$(jq -r '.runId' mission-admission.json)"
praxa mission get --run-id "$RUN_ID" > mission-current.json
jq '{runId, status, sequence, steps}' mission-current.json

Poll conservatively or use the SDK event stream for a real-time application. Do not run an unbounded tight loop from CI.

for attempt in 1 2 3 4 5; do
  praxa mission get --run-id "$RUN_ID" > mission-current.json
  STATUS="$(jq -r '.status' mission-current.json)"
  case "$STATUS" in
    succeeded|failed|cancelled) break ;;
  esac
  sleep 2
done

Request cancellation

CANCEL_KEY="cancel:$(uuidgen | tr '[:upper:]' '[:lower:]')"

praxa mission cancel \
  --run-id "$RUN_ID" \
  --reason "Operator ended the disposable acceptance test" \
  --idempotency-key "$CANCEL_KEY" \
  > cancellation.json

Cancellation is a request. Read the mission again until its authoritative status is terminal. Do not treat the command’s successful HTTP response as proof that an in-flight provider operation was rolled back.

Automate safely

set -euo pipefail

result="$(praxa mission create \
  --input mission.json \
  --idempotency-key "$IDEMPOTENCY_KEY")"

run_id="$(jq -er '.runId' <<<"$result")"
status="$(jq -er '.status' <<<"$result")"
printf 'admitted run=%s status=%s\n' "$run_id" "$status"

Check the process exit before parsing stdout. Send diagnostics to stderr and avoid printing environment variables.

End-to-end test

Use a disposable principal with only the required mission scopes.
Create one mission and persist its key, body digest, and run ID.
Replay the exact request and require the same logical mission.
Change the body under the old key and require a conflict.
Read the run to a terminal state or request cancellation.
Repeat with an under-scoped token and require denial.
Revoke the token and require the next command to fail.
Remove disposable files and revoke temporary credentials.

Command reference

Review every mission flag, bound, output, and required environment value.

Mission lifecycle tutorial

Add resumable SSE, signals, abort handling, and terminal reconciliation in TypeScript.

Workflows

Automate and operate

Run mission workflows with the Praxa CLI

Choose intent or mission intake

Submit intent

Create a budgeted mission

Read authoritative state

Request cancellation

Automate safely

End-to-end test

Command reference

Mission lifecycle tutorial

​Choose intent or mission intake

​Submit intent

​Create a budgeted mission

​Read authoritative state

​Request cancellation

​Automate safely

​End-to-end test

Command reference

Mission lifecycle tutorial

Choose intent or mission intake

Submit intent

Create a budgeted mission

Read authoritative state

Request cancellation

Automate safely

End-to-end test