m9m terms, defined.

The vocabulary you'll meet in the docs, guides, and comparisons — in plain language.

Workflow

A directed graph of nodes that m9m executes. m9m imports n8n workflow JSON unchanged, so the same node names and connections carry over.

Node

A single step in a workflow. m9m ships 32+ built-in node types (HTTP, databases, transforms, control flow, cloud, comms) and lets you write custom nodes in Go.

Trigger

What starts a run. m9m supports cron schedules and HTTP webhooks natively, with no separate scheduler process.

CLI node

A node that launches an external command-line agent — Claude Code, Codex, or Aider — inside a sandbox with CPU, memory, and network limits.

MCP server

The built-in Model Context Protocol server. It exposes 37 workflow-management tools so an MCP-aware LLM can list, edit, run, and debug workflows.

Sandbox

The Linux-namespace isolation m9m uses to run agent CLI nodes, with cgroup CPU/memory limits and network allow-lists so an agent cannot touch the rest of the host.

Expression

The n8n-compatible templating syntax — {{ $json.field }}, {{ $node["name"].data }} — that pulls data from earlier nodes into the current one.

Checkpoint

A resumable point in a run. Combined with human-in-the-loop steps, checkpoints let a workflow pause for approval and continue where it left off.

Audit log

The append-only record of every run and agent action. Included in the open-source release, not reserved for a paid tier.

Versioning

Git-based history of every workflow edit, so you can diff, review, and roll back changes to a workflow.

New to m9m? Start with how it works, then browse the use cases or the full FAQ.