Claude and ChatGPT

Claude.ai and ChatGPT can both connect to Testsnag. Once connected, you can ask in plain language which tests failed overnight, run your release suite, or have the assistant explain a failure and fix the test. It works with the same tests, runs and results you see in the dashboard, in the browser, the desktop apps and on your phone.

Using a coding assistant such as Claude Code, Codex, Cursor or VS Code instead? See AI clients (MCP).

Connecting Claude

Testsnag is listed in the Claude connector directory, so you can add it in a couple of clicks.

1. 1

   Open connectors

   In Claude, open Customize, then Connectors.

2. 2

   Find Testsnag

   Open the Discover tab and search for Testsnag.

3. 3

   Connect, sign in and approve

   Choose Connect. Claude opens Testsnag so you can sign in, pick a workspace and approve. You are then sent back to Claude.

Prefer to add it yourself? Choose Add, then Add custom connector, name it Testsnag and use the server address https://testsnag.com/mcp.

Claude lists the Testsnag tools as read-only or write, and you choose whether each group runs straight away or asks you first. On Claude Team and Enterprise plans, an owner may need to add the connector for the organization first.

Connecting ChatGPT

ChatGPT calls connectors plugins. Adding your own MCP server needs Developer mode, which you can turn on in ChatGPT settings. On Business and Enterprise plans, a workspace admin may need to allow it.

1. 1

   Open plugins

   In ChatGPT, open Plugins from the sidebar.

2. 2

   Add a custom MCP server

   Choose Add, then Add custom MCP server. Name it Testsnag, paste https://testsnag.com/mcp as the server URL and keep OAuth as the authentication. Tick I understand and want to continue, then choose Create as a plugin.

3. 3

   Sign in and approve

   Choose Connect Testsnag. ChatGPT opens Testsnag so you can sign in, pick a workspace and approve.

4. 4

   Use it in a chat

   In a new chat, choose Plugins under the message box and pick Testsnag.

What you approve

When you approve, you choose one workspace and either Read only or Read and write access. With Read only, the assistant can look at tests, runs and results but cannot change anything or start a run. The connection acts with your role in that workspace, so it never sees more than you can.

The connection appears as a connected app on the API tokens tab of your workspace settings. Delete it there to disconnect straight away. To use a different workspace, connect again and pick it.

Things to ask

• "Which of my tests failed in the last day, and why?"
• "Run the Release smoke suite and tell me when it finishes."
• "Why did the last run of Checkout, guest user fail? Suggest a fix."
• "Create an API test that checks GET https://example.com/health returns 200 in under a second, then run it."
• "Add a step to the sign up test that checks the welcome email arrives."
• "Schedule the Homepage SEO audit to run every morning at 7."
• "How many runs do I have left this month?"

Testsnag tells both apps which actions only read, which change things, and which are destructive, such as deletes, overwrites and revoking a link or secret. The apps use this to decide when to ask you before going ahead: by default ChatGPT runs low-risk actions straight away and asks before the rest.

What chat apps cannot do

Claude.ai and ChatGPT cannot read files on your computer, so they cannot upload a mobile build. Upload it in the dashboard or with the command line tool, then ask the assistant to attach it from your Files library.

Never paste passwords into the chat. Add them as secret variables on the Variables page and tell the assistant the name. The assistant can use a secret in a step but can never read its value.