ARCY MCP
Connect your coding agent once for the whole workspace, run one prompt that sets a Product up, proposes Specialists or battle tests your chat, and decide what it may read.
Your coding agent sets your Products up from your repository. You connect it on Workspace settings > ARCY MCP, once for the whole workspace.
Before you start
- Only an Admin presses Allow when an agent signs in, and only an Admin revokes one.
- One sign-in reaches every Product in the workspace, so each call names its Product.
- Point your agent at a local or staging copy: a preview calls your API for real.
- An agent reads and proposes; it cannot publish, delete or accept.
Setting it up
Connect your agent. Pick your client. No key goes in any command or file, and the address carries no Product, so adding one later changes nothing. Or press your agent's logo: it copies a task that has the agent add the ARCY MCP itself, and you only sign in.
Allow the sign-in. Allow only a sign-in you started; the card marks an unverified client, names how many Products the agent reaches, and asks nothing else.
Connected agents. Every agent signed in to the workspace. One token per client per person; signing in again replaces it and keeps its open proposals.
Let a connected agent read visitor behaviour. Off until an Admin turns it on, which records who and when. While it is off your agent reads your setup and never what visitors did, so Propose Specialists from your code and real behaviour works from your code and setup alone and says so. What it may read once on is in What ARCY collects.
What your agent can do. One row per Product opens its entries, each line already naming the Product. Set up ARCY end to end, marked Start here, covers one Product per run, then names the rest. It updates an old arcy.js, making each code change the Changelog lists, then runs all twelve surfaces, each also its own entry.
Propose Specialists from your code and real behaviour. Your agent reads the Product's setup and, with the switch above on, where sessions stopped, then answers with a plan; one confirmation leaves drafts waiting on Specialists.
Battle test the chat and fix what it finds. Your agent writes conversations a visitor of your Product would have, ARCY plays each against your Development chat, and your agent proposes only the fixes a re-run proved. Every run stays on Persona. It calls your public operations for real, which How AI request content is handled covers; it writes nothing.
How it asks. Which Product a repository is for is yours to say, in every mode, as is anything only you know, such as your legal name. Workspace answers carry to the next Product.
What it changes
Every write outside a flow draft is a proposal, on the Product the call named, and the page that owns it is where Accept applies it with that page's checks. Flows land as drafts on Flows; imported rows wait as an Import. A Domains or spend cap proposal is the workspace's and waits, one at a time, on Domains or Limits. Revoke discards open proposals and names each Product it discards from; accepted work and the flows written are kept.
Four prompts edit your repository, none touching git. Install arcy.js wires the Quickstart code and owns ARCY_TOKEN, ARCY_SECRET, ARCY_DEV_TOKEN and ARCY_DEV_SECRET, never a Secret's value; with Update arcy.js, it gives an npm install its Update bot. Set up your events writes your arcy.track() calls against Events. Update your policies writes the Legal templates into your own. Remove ARCY takes it all out and clears returning visitors' storage.
Every tool but six names its Product; the workspace reads and the Domains and spend cap proposals name none, and one agent makes at most 180 calls a minute. Not on the MCP: file uploads, Require identity verification, DNS records, deleting a Product, and Accept itself.
When it goes wrong
- The agent gets 401 or says it needs login. Sign in again from the agent (in Claude Code,
/mcp, then Authenticate). - A call is refused for naming no Product, or no Product to name is on offer. Your agent holds a tool list from before this changed: reconnect it. If a name is refused as not found, check it against the slug in the dashboard URL and that the Product is not Not serving.
- The card says the link has expired. Ten minutes passed, or another tab answered it: start again from your agent.
- The card has no Allow. You are a Member: an Admin signs in from their own agent.