Skip to main content

Atomic Executor

JARVIS Atomic Executor executes atomic NodeRuns (leaf work).

It is intentionally stateless: it does not persist Run/NodeRun state. The Run Coordinator is the system-of-record.

Spec vs implementation

For the normative API contract, see ARP Standard: Atomic Executor.

Reference implementation​

  • GitHub: AgentRuntimeProtocol/JARVIS_AtomicExecutor
  • PyPI: arp-jarvis-atomic-executor
  • GHCR: ghcr.io/agentruntimeprotocol/arp-jarvis-atomic-executor:0.3.3

In the recommended JARVIS_Release Docker Compose stack:

  • Compose service: atomic-executor
  • Internal URL: http://atomic-executor:8082
  • Exposed host URL: (not exposed by default)

Running outside Docker (advanced)​

Use the single meta CLI arp-jarvis (preferred over component CLIs):

STACK_VERSION="0.3.3" # match `JARVIS_Release/stack.lock.json`
CLI_VERSION="0.3.8"   # meta CLI release
python3 -m pip install "arp-jarvis==${CLI_VERSION}"
arp-jarvis atomic-executor -- --host 127.0.0.1 --port 8082

Node packs​

In JARVIS v0.x, atomic handlers are loaded in-process from installed node packs via Python entry points:

  • entry point group: jarvis.nodepacks

The first-party pack is:

  • GitHub: AgentRuntimeProtocol/JARVIS_AtomicNodes
  • PyPI: arp-jarvis-atomic-nodes