Foundation~25 minHazelJS

Hazel CLI Agent Ops

Use hazel agent doctor, run, logs, and runs — and know DNA smoke stubs are not the same as app runtime.execute.

Authors: editorial-team
Published: Published 2026-08-01
Last reviewed: Reviewed 2026-08-14

Progress is stored locally in this browser.

Direct answer

@hazeljs/cli helps you doctor the environment, smoke-test DNA, read timelines, and list durable runs. Real tools and HITL continuation still happen in your app via AgentRuntime.execute / approveAndResume.

Docs: CLI. Guide: /guides/design-an-agent-manifest.

Critical distinction (do not skip)

Path	What it is
`hazel agent run ./agent.dna.json "…" --mock`	DNA bootstrap with stub tools
Your app `runtime.execute('ops-desk', …)`	Real `@Tool` handlers
`hazel agent runs resume\|approve` (file store)	Records HITL intent; app must still `approveAndResume`

Never demo “production behavior” using only DNA stub smoke.

Step 1Daily local ops loop

bash

# Environment / wiring checks
hazel agent doctor

# Optional DNA smoke (stubs) — offline LLM with --mock
hazel agent run ./agent.dna.json "Status of ORD-1001?" --mock

# Timeline / logs (path depends on your starter)
hazel agent logs --timeline .hazel/runs/timeline.jsonl

# Durable runs on disk
hazel agent runs list --dir .hazel/runs

Step 2When to use which

Situation	Use
New laptop / CI image	`doctor`
Manifest / DNA sanity	`hazel agent run` smoke
Tool truth & product QA	App execute + goldens
Incident “what happened?”	`logs --timeline` + Inspector
Approval after suspend	App `approveAndResume` (CLI may mark store intent)