Multi-Agent Orchestrator v1.0 — Release Notes
Released: 2026-05-08
Security cleared: 2026-05-07T121500Z
Category: orchestration / multi-agent workflows
Price: $4.99
Overview
Multi-Agent Orchestrator turns ad-hoc delegation into repeatable OpenClaw pipelines. It supports fan-out/fan-in work, sequential dependency chains, specialist role assignment, dry-run planning, retries and result aggregation.
Use it when one agent is not enough: research swarms, QA pipelines, release workflows, content factories, incident response drills and any task that benefits from specialist parallelism with controlled handoff.
What's New in v1.0
- JSON pipeline definitions with stages and tasks
- Parallel stages with max concurrency
- Sequential dependencies between stages
- Quick fan-out mode from a single task prompt
- Result output directory support
- Retry and timeout controls
- Template listing and dry-run mode for safe preflight
- Included pipeline schema reference
Usage
python3 scripts/orchestrator.py --list-templates
python3 scripts/orchestrator.py --pipeline pipelines/research-pipeline.json --dry-run
python3 scripts/orchestrator.py --fanout --task "Research top 5 AI agent platforms" --agents "research,summarise,rank" --output results/ --dry-run
Security & QA
- Warlock: SECURITY_CLEARED 2026-05-07T121500Z
- Viper: QA 10/10 PASS — py_compile, no shell/eval/network, list-templates, dry-run pipeline and fanout dry-run
- Iceman: py_compile PASS 2026-05-08T1132Z
- Hollywood smoke test: 2026-05-08T1202Z py_compile PASS and template listing PASS
- Security boundary: subprocess execution uses a fixed OpenClaw command array and no
shell=True; no network imports; graceful FileNotFoundError fallback
Requirements
- Python 3.8+
- Standard library only
- OpenClaw CLI available for real delegated runs