Stomio
AI Help

Connect Your AI Assistant to Stomio

Aug 31, 2026·7 min read

You already have an AI assistant open most of the day. Connecting it to Stomio means it stops answering from general knowledge about beta testing and starts working on your program: your testers, your tickets, your results, your announcements.

This guide covers connecting an assistant, choosing what it may do, and turning it off again.

In this article

What does connecting an assistant actually do?

Stomio is a remote MCP server. MCP is an open standard for letting an AI assistant work with a system that holds real data, and Stomio publishes more than eighty tools through it, covering beta projects, tasks and steps, tickets, testers, announcements and applicant screening.

Once connected, your assistant can read your beta data to answer a question, and, if you allow it, do the work as well: draft a task set, reply to a ticket, send an announcement, shortlist applicants. It is the same platform with the same rules, reached from a conversation instead of the manager view.

What do I need before I start?

  • A manager or owner account in the workspace. Your assistant gets exactly the permissions you have, never more.
  • AI features enabled on the beta projects you want it to reach. That switch lives on the beta project itself, and a beta with AI features off is invisible to an assistant.
  • An assistant that supports remote MCP servers, such as claude.ai, Claude Desktop, Claude Code or ChatGPT.

You do not need an engineer, an API key or anything installed. The connection is a sign-in, like connecting any other account.

How do I connect my assistant?

1. Copy your connect URL. In Stomio, open Workspace settings and go to AI tools. The connect URL is at the top, with a Copy button next to it. It is the same URL for every assistant you connect.

2. Add it to your assistant. Every client words this differently, but they all have one place to add a connector or an MCP server: look for connectors, integrations or MCP settings, choose to add a custom or remote server, and paste the URL.

3. Sign in and choose what to allow. Your assistant sends you to Stomio to sign in. The consent screen that follows is where you decide how far the connection goes: reading your beta data only, or reading and doing the work as well. Approve it and the connection is live, usually within a minute of starting.

What you choose there is fixed for that connection. To widen it later, or to pull it back, connect the assistant again and answer the consent screen differently. The new choice replaces the old one, so there is never any doubt about what a connection is allowed to do.

What if my assistant asks about authentication or client settings?

Most assistants need nothing but the connect URL and your sign-in. A few show a setup panel with extra choices first. None of them need thinking about, but two of them will quietly stop the connection working if you pick the wrong one, so here is what each is asking and what to choose.

Authentication: choose "Always required". Stomio asks every assistant to sign in before it can use any tool, so there is nothing to reach without an account. Choosing "None" leaves your assistant with no way in, and the connection will not work.

OAuth client: choose "No client ID, register one automatically". Stomio registers your assistant for you when you connect, which is why there is nothing to fill in. Watch this one, Your assistant may offer to hand Stomio the client details it hosts itself, which claude.ai calls "Use Anthropic's hosted client metadata" and marks as recommended. Stomio does not read client details from a URL your assistant hosts, so picking it stops the sign-in with an error.

Transport: leave it as "Streamable HTTP". Your assistant fills this in from the connect URL. Stomio does not accept SSE.

Other assistants word these differently, and some ask none of them. Wherever they turn up, the answers are the same: sign-in required, let the assistant register itself, and leave the transport your assistant picked from the URL.

What can I allow it to do?

Reading is granted when you connect: your beta projects, tickets, testers, workspace members and the announcements you have already sent.

Doing things is opt-in, one area at a time, so you can start with an assistant that only reads:

  • Announcements. Send announcements to your testers.
  • Testers and applicants. Invite, label and remove testers, re-send invitations, and manage applicants.
  • Tickets. Create and update tickets, reply to them, and manage their assignees and tags.
  • Tasks and steps. Create, edit, delete and reorder the tasks and steps in a beta.
  • Beta projects. Create and edit beta projects and manage their lifecycle: publish, launch, pause, archive, restore.

A connection that was granted reading only cannot write anything, whatever it is asked to do. Changing that means reconnecting the assistant and approving the wider access on the consent screen.

What should I ask it first?

Good openers, in the order most managers find useful:

  • "Summarize what my testers reported this week and what I should fix first."
  • "Which tasks are testers dropping out of, and what do they say when they do?"
  • "Find every tester on firmware 2.1 who reported a connection problem."
  • "Draft the week-two task set based on the program we ran on the last version."
  • "Triage the tickets that came in overnight and tell me which ones block the release."
  • "Write the mid-beta announcement, mention the two fixes that shipped, and send it to the testers who finished week one."

The last two are worth trying once your assistant has read access working, because they show the difference between an assistant that describes your beta and one that runs errands inside it.

What will it never do on its own?

Anything that would reach a tester or an applicant, an announcement, an invitation, a ticket reply, is previewed in the conversation first and does nothing until you confirm it. You see who it would reach and what it would say before it happens.

Beyond that, three limits hold whatever you or the assistant intend. It only reaches beta projects with AI features enabled. It only has your own permissions. And email addresses, phone numbers and precise locations are stripped before results leave Stomio, except where an email address is the only identifier a person has, which is the case for applicants and for people whose invitation is still pending. There is no tool that exports your contact list.

Does this use my AI credits?

No. AI credits cover the AI features inside Stomio, such as AI Canvas and the AI insights buttons. Work you do through your own assistant runs on that assistant's model and subscription, so it does not touch your Stomio credits.

How do I disconnect an assistant?

Open Workspace settings, then AI tools. Every connected assistant is listed with what it was allowed to do and when it was connected. Disconnect ends its access immediately, and it has to go through the whole consent flow again to come back.

Disconnecting is per assistant, so removing one leaves the others working.

Why can my assistant not see one of my betas?

Almost always because AI features are off for that beta project. Open the beta, turn AI features on, and ask again.

If that is not it, check that you can see the beta yourself in the manager view. The connection carries your permissions, so a beta you do not manage is not one your assistant can reach either.