Connect Your Agent to Console Testers with MCP

announcementsguidesmcp

18 Oct '26by Console Testers

We built an MCP server for Console Testers, so any MCP-capable agent (Claude, Cursor, or your own) can check your closed test, add apps, and pull reports without you opening the dashboard.

What MCP actually is

MCP stands for Model Context Protocol. Strip away the spec and it is a standard way for an AI app to call tools that live on a remote server. Your agent finds out what tools exist and what arguments they take, then calls them like it would call anything else.

We built one because half of running a closed test is checking things. How many days are left, whether the report is ready, if you can submit yet. You ask your agent those questions anyway, so we made it able to answer them.

What your agent can do

Eight tools, in two groups.

The read-only ones your agent can call whenever it likes:

`check_token_balance` - your balance, plus a nudge to buy more when it hits zero
`check_app_can_be_added` - free dry run covering package format, duplicates, store listing, and token cost against your balance
`get_all_apps` - every tracked app with its platform, plus the balance
`check_app_status` - dashboard URL, progress, status, days remaining, report readiness
`can_submit_review` - whether a review can go in, and whether one already exists
`get_report` - whether the two testing reports exist, with links to both

The other two change things:

`add_app_for_testing` - adds an app by package name and platform, and spends tokens
`submit_review` - submits or updates your review, which then goes to moderation

Those two are the only tools that write, and the setup instructions we generate tell your agent to show you the exact details and wait for your explicit confirmation before calling either one. Nothing spends your tokens without you saying yes.

Setup

Open your dashboard and go to the MCP page. You get two things there: your active connections, and a paste-ready block under "Agent self-setup".

Copy that block into your agent. It carries the server config plus a few instructions, so there is nothing for you to configure by hand. The config itself is this:

json

{
  "consoletesters": {
    "type": "http",
    "url": "https://consoletesters.com/api/mcp"
  }
}

No API key, no header to paste around. The server uses standard MCP authorization (OAuth 2.1), so the first connect opens a browser page where you sign in to Console Testers and approve access. Your agent holds onto those tokens and refreshes them itself.

Once it is connected, ask it to list the server's tools. You should see all eight above, and then you are done.

Your connections stay listed on the same dashboard page with when each client was added and when it last ran. Disconnect one and that agent is locked out until you sign it back in.

Prompts worth trying

"How is my closed test going?"
"Which of my apps can I submit a review for right now?"
"Are the testing reports ready for com.example.app?"
"Do I have enough tokens to add another app?"

The agent calls the read tools, gets real data back, and answers. No tabs, no copy-pasting package names.

Why bother

The dashboard works fine. This is about how many times you check it over fourteen days. Status questions get answered in chat, you stop mistyping package names, and you can ask for a summary across every app in one go instead of clicking into each one.

If you already live in an agent all day, this is one less reason to leave it.

by Console Testers