Skip to main content
Automate It ships as an AgentSkills-format skill: a SKILL.md the agent reads plus ait.mjs, a zero-dependency CLI (Node 18+) that speaks Automate It’s MCP server. The skill teaches the full loop — delegate a task, or generate content yourself and submit it through the human review gate.

Install

The skill is published on ClawHub:
(clawhub install @workingdevshero/automate-it works too, and the skill folder can always be dropped into your workspace’s skills/ directory by hand.)

Get an API key

The skill authenticates with a scoped ak_… key — there’s no OAuth flow for this path, so create the key first:
  1. In the Automate It web app, click your avatar at the bottom of the sidebar to reach Profile (https://automate.it.com/app/<workspace-slug>/settings/profile). Keys can’t be minted from the mobile apps.
  2. Under Developer API Keys, click Create Key.
  3. Name it something you’ll recognize (“OpenClaw agent”) and tick tasks:read, tasks:write, content:write, and skills:read — enough to create tasks, submit content into review, and read your workspace’s voice and formatting skills.
  4. Copy the ak_… secret. It’s shown once — hold it somewhere safe until you store it below, and don’t paste it into a chat.
Add files:read if the agent reads skill-bundled reference files and files:write if it uploads media. The full table is in API keys & scopes.

Configure

Install the CLI the skill drives, then give OpenClaw the key:
The skill declares primaryEnv: AUTOMATE_IT_API_KEY, so skills.entries.automate-it.apiKey accepts a secret reference. Keep the value in OpenClaw’s shared secret store and reference it — the key never lands in openclaw.json:
store set takes the value on stdin or at a no-echo prompt — it refuses --value so the key stays out of shell history. To keep credentials in the Gateway environment instead, use --ref-source env and put the value in the Gateway env file at mode 600 (~/.openclaw/.env, or ~/.openclaw/gateway.systemd.env on a Linux systemd install), then openclaw gateway restart.
Don’t set skills.entries.automate-it.env.AUTOMATE_IT_API_KEY. That map holds literal strings, rejects secret references, and leaves a live publishing credential readable in config to anything that can read files — openclaw doctor flags it as plaintext secret-bearing config. And never paste the key into a chat message: if the agent asks for it, have it request the value through OpenClaw’s secrets tool, which prompts you out of band and stores it without the key entering the transcript.
AUTOMATE_IT_WORKSPACE is optional and not secret — set it in skills.entries.automate-it.env if the key can see more than one workspace. If your agents run sandboxed, pass the key into the sandbox as well (agents.defaults.sandbox.docker.env): skill apiKey and env injection applies to host runs only, and secret-store values don’t cross that boundary either.

Verify

From here the skill’s own instructions take over: the agent creates tasks, works assigned ones, revises after rejection, and fetches published links. The Worker mode guide is the human-readable version of that loop, and the CLI reference covers every command.