Antwork
Start free
How it worksPricingSign in
LANGUAGE
[Setup · one prompt]

Set up Antwork from your AI

Paste one sentence into the AI you already use. It reads this page, installs Antwork where it can, walks you through signing in and checks the connection. You approve every step, and nothing is posted during setup.

[Paste this into your AI]
Read antwork.io/setup and set up Antwork for me.
[What your AI does]
  1. 01
    Checks where it is runningWhich AI and version, and whether Antwork is already connected. It keeps your other tools as they are.
  2. 02
    Installs AntworkClaude Code gets the plugin, with the Antwork skills. Other agents add the server at api.antwork.io/mcp. In Claude and ChatGPT it tells you the clicks to make instead.
  3. 03
    Signs you inYour browser opens Antwork’s sign-in. There is no API key and nothing to paste into the chat.
  4. 04
    Checks the connectionIt reads your workspace, a free and harmless check, and only says it works once that read succeeds.
  5. 05
    Hands you the next stepConnect your first social account, then ask for a week of posts. It suggests; you decide.
Sign in, no keysOAuth in your browser. There is no API key to paste into a chat, a file or a repository.
Nothing posts during setupYour AI only reads to check the connection. Every post later starts as a draft you approve.
Free to startThe Free plan has 2 accounts and 10 posts a month, and every new account starts with a 7-day Pro trial. No card.
[For AI agents]

The instructions your AI reads

Shown in full so you can see exactly what you are asking it to do. Agents can also fetch the plain text at antwork.io/setup.md.

setup.md
Set up Antwork in this agent. Do what you can; guide me through anything that needs my input.

## 1. Check this agent

Identify this agent and its version. Ask only if you cannot tell.
Check whether Antwork is already connected — an MCP server named `antwork`, or Antwork skills already loaded. Preserve my other integrations and do not add a duplicate.

## 2. Install the plugin first

On Claude Code, the official plugin is one step for both halves: it wires up the MCP server and loads the Antwork skills, which teach the correct tool order, the draft-then-publish two-step and the per-platform limits.

/plugin marketplace add polynode-co/antwork
/plugin install antwork@antwork

Other agents: there is no plugin. Go to step 3.
Check the installed client's own help before running commands.

## 3. Fall back to the MCP server alone

Add `https://api.antwork.io/mcp` as a streamable HTTP MCP server named `antwork`.
Use this agent's native command or config file, looked up in its current documentation — do not guess a config schema. Reference: https://antwork.io/docs/mcp
The skills are Claude Code only. Do not hand-copy them into another agent; without them the tools still work, and https://antwork.io/docs/mcp has the workflow each one encodes.

## 4. Sign in

OAuth only. The MCP server issues no API keys, so there is no key to paste and nothing belongs in chat, in a repository or in an environment variable.
Start the flow and let me approve it in my browser. The consent screen signs me into my Antwork account.
I need a free Antwork account first — https://antwork.io/login. MCP access is not gated on a paid plan.

## 5. Reload and verify

Use this agent's native reload flow, checking its installed version, help or official documentation. Prefer automatic discovery or an in-place reload; restart only if it is required to load the new tools.
Once the tools load in this session, run `list_workspaces`. It is a free read. Check skill discovery too, if you installed the plugin.
Track installation, sign-in and verification separately. A registered server is not proof that this session can call its tools. Never claim verification before that read succeeds. If a reload needs my action, give me the instruction rather than retrying unavailable tools.
Do not create, schedule or publish anything during setup. Antwork's write tools reach real social accounts, so there is no harmless demo post.

## 6. Finish with a short handoff

Keep progress updates brief. The final reply must be 140 words or fewer and follow this template:

**Status**
[Briefly say what succeeded or what blocked setup.]

**Next**
1. [Give the native reload command or UI action for this agent, only if needed.] Then say "Check that Antwork is connected." Approve sign-in if prompted.
2. Try one of these:
   - [Setup invocation] **(recommended)** — connect a social account and set your timezone and posting times.
   - [Ideas invocation] — post ideas grounded in what already performed on your accounts.
   - [Post invocation] — draft a post in your own voice, then schedule it.
   - [Audit invocation] — score your current social presence and say what to fix.

Want to know what was set up? Just ask.

Adapt the template to the actual result:
- If installation failed, name the blocker and replace the reload line with the fix. If fully verified, say it is ready and omit the reload. Do not claim sign-in failed or is required merely because tools need reloading.
- Recommend this agent's idiomatic way to invoke each installed skill: its native command, mention, picker or plain request. With the plugin these are `/antwork setup`, `/antwork ideas`, `/antwork post <idea>` and `/antwork audit`; do not assume that syntax works in an agent without it — there, give the equivalent plain-language request against the connected tools. Recommend workflows; do not run them during setup.
- Keep tool names such as `list_workspaces` in your checks, not in the final reply. Omit versions, paths, connection details, skill counts, other integrations and cleanup commands unless they explain the blocker or I ask. Do not add sections or a verification checklist.