AI agent orchestration
Run Claude Code, Codex, and Aider as sandboxed steps inside the same workflow that handles your webhooks, HTTP calls, and databases.
The problem
Coding agents are powerful but unbounded: given a shell they can touch anything on the box, and there is no record of what they did. Bolting a separate agent stack next to your automation tool doubles the infrastructure and still leaves the sandboxing to you.
How m9m does it
m9m treats agent orchestration as first-class. CLI nodes launch Claude Code, Codex, or Aider inside Linux namespaces with CPU, memory, and network limits, and every action is written to the audit log. The same workflow can gate the agent behind a human-review step and re-run it on failure.
How it works, step by step
- 1
Add a CLI node
Drop a cli.claude-code (or cli.codex / cli.aider) node into your workflow and set sandbox, cpu, memory, and network limits.
- 2
Feed it context
Pipe upstream node output — a PR diff, a failing test log, a ticket — into the agent prompt with the {{ $node.name.field }} expression syntax.
- 3
Gate and audit
Add a human-review step before any irreversible action. Every run is versioned in Git and written to the audit log.
Nodes involved
Keep reading
Need help shipping agents or migrating off n8n?
Neul Labs — the team behind m9m — takes on a limited number of consulting engagements each quarter. We help teams migrate n8n workflows, build custom Go nodes, sandbox AI agents in production, and design automation platforms that don't collapse under load.