Connection guides · Connection guide

ChatGPT and TestingOps: release status, cases and runs in a ChatGPT conversation

Add TestingOps to ChatGPT as a custom MCP server and ask whether a release is ready, whether a case already exists or what changed between two runs, and, where your ChatGPT plan allows writes, have ChatGPT create cases and start runs.

Connection guide
Setup guide

Request a demo OpenAI: connect a custom MCP server to ChatGPT

ChatGPT RequestResult TestingOpsAI test agents

Connection: Connection guide
Writes back: Depends on the access you grant
Last checked: 7 Nov 2026

In short

ChatGPT and TestingOps in short

ChatGPT can use TestingOps when the TestingOps MCP endpoint, which ends in /api/mcp/v1, is added to ChatGPT as a custom MCP server; ChatGPT then calls TestingOps tools in the conversation.

ChatGPT connects to TestingOps with sign-in: the person approves it on a TestingOps consent page and chooses the company or customer and read only or read and write access.

OpenAI decides which ChatGPT plans can add custom MCP servers and write through them, and in ChatGPT Business and Enterprise workspaces an admin creates and publishes the connector.

ChatGPT reaches the endpoint from OpenAI's side, so the TestingOps address must be reachable from ChatGPT over https.

What it does

What it does for your testing

1
Release readiness with its gaps
Input
A question such as "is this release ready?"
TestingOps
Assesses open blockers, quarantined tests, cases that never ran and coverage
Output
A recommendation with an explicit list of what is not covered; the decision stays with people
2
A duplicate caught before it is written
Input
The title and draft steps of a case someone plans to add
TestingOps
Checks them against the existing cases and the knowledge base
Output
The closest existing case to reuse, or confirmation that the case is new
3
Two runs compared in plain language
Input
Yesterday's run and the one before it
TestingOps
Compares them for regressions, fixes and cases that still fail
Output
The cases that passed before and fail now, listed first

Step by step

How it works

Copy the endpointThe MCP server card under Integrations → API & MCP shows the Endpoint (streamable HTTP); the address ends in /api/mcp/v1.
Add a custom MCP server in ChatGPTPaste the address where ChatGPT adds custom MCP servers and choose sign-in. In Business and Enterprise workspaces an admin creates it and publishes it to members.
Approve on the TestingOps consent pageChatGPT registers itself; you choose the company or customer and read only or read and write. The page names the address the browser returns to.
Ask in a conversationChatGPT calls TestingOps tools when a question needs them, as the person who approved it.
End access when you need toRemove ChatGPT under Connected apps; it loses access immediately.

OpenAI names and places ChatGPT's settings and decides which plans can use custom MCP servers; follow OpenAI's guide for the ChatGPT side.

Connect

Setup

In TestingOps, open Integrations → API & MCP and copy the Endpoint (streamable HTTP).
In ChatGPT, add a custom MCP server with that address and choose sign-in (OAuth). In Business and Enterprise workspaces an admin does this and publishes it to the workspace.
When ChatGPT opens the TestingOps consent page, pick the company or customer and read only or read and write, then approve.
Check the approval under Connected apps; remove it there to end access.

Documentation: OpenAI: connect a custom MCP server to ChatGPT

Where you connect it in TestingOps: Integrations → API & MCP → Connection guides → ChatGPT

Use cases

Go/no-go preparation

Trigger: A product owner prepares a release meeting in ChatGPT
Uses: Release readiness: open blockers, quarantined tests, never-run cases and coverage
Agent activity: ChatGPT reads the assessment and lists the gaps next to the recommendation
Result: A readiness summary that states what was not tested
Condition: Read only access is enough; people make the release decision

Check before writing a new case

Trigger: A tester drafts a case with ChatGPT
Uses: Duplicate check against the project's cases and knowledge base
Agent activity: ChatGPT checks the draft before suggesting it and points to an existing case when there is one
Result: Fewer duplicate cases in the inventory

Cases from a requirements list

Trigger: An analyst pastes a list of requirements and asks for cases
Uses: Bulk case import into a project
Agent activity: ChatGPT writes the cases and imports them into the chosen project in one call
Result: Cases in TestingOps instead of a spreadsheet attachment
Condition: Needs read and write access in TestingOps and a ChatGPT plan that allows writes

Capabilities

Capabilities and limits

What is read

Release readiness, coverage and traceability for the projects the signed-in person can open
Existing cases and the knowledge base, including duplicate checks for a draft case
Runs, comparisons between runs, results and captured evidence
Defects and their status

What is produced

ChatGPT's answers and summaries, built from TestingOps tool results; readiness answers include what is not covered

Where it can write

Only when the consent was read and write and the ChatGPT plan allows writes: cases and scenarios, case imports, runs, defects, data sets and items in trackers connected to TestingOps.

How it connects

A custom MCP server in ChatGPT pointing at the TestingOps MCP endpoint, with OAuth sign-in; ChatGPT registers itself with TestingOps

Account and permissions

Sign-in on the TestingOps consent page: company or customer, read only or read and write
In ChatGPT Business and Enterprise workspaces, an admin creates and publishes the connector
The connection acts as the person who approved it; a customer account sees only that customer's projects

Limits

The TestingOps guide for ChatGPT uses sign-in; access keys are meant for tools that send a fixed header
OpenAI decides which ChatGPT plans can add custom MCP servers and which can write through them
The TestingOps address must be reachable from ChatGPT over https
Tools only: no MCP resources or prompts

Example

Example: how a team uses it

Illustrative scenario, not a customer story.

Before a release meeting, a product owner asks ChatGPT whether the release is ready. ChatGPT reads the TestingOps readiness assessment: one blocker is still open, a test is quarantined and a group of cases in the release scope has never run. It lists those gaps next to the recommendation instead of giving a plain yes or no.

The team decides to run the missing cases first. Because their workspace allows writes and the product owner approved read and write, ChatGPT starts the run on the staging environment, and the owner checks the result in the same conversation later that day.

Frequently asked questions

Does ChatGPT use an access key?

The TestingOps guide for ChatGPT uses sign-in: ChatGPT registers itself and you approve it on the TestingOps consent page. Access keys are for tools that send a fixed header, such as editors and Claude Code.

Which ChatGPT plans can use TestingOps?

OpenAI decides which plans can add custom MCP servers and which can write through them. At the time of writing, writing needs a Business, Enterprise or Edu workspace and Pro is read only; check OpenAI's guide for the current rules.

Who sets it up in a Business or Enterprise workspace?

A ChatGPT admin creates the custom MCP server and publishes it to the workspace. Each person who uses it approves TestingOps access with their own account, so their own permissions apply.

How long does an approval last?

The consent gives ChatGPT a 30-day access token, which it renews with a refresh token valid for 180 days. Removing ChatGPT under Connected apps ends access at once.

Will ChatGPT offer tools that a customer account cannot use?

No. Customer accounts get a shorter tool list, without the tools meant for TestingOps' own agents, source code and system logs. TestingOps also checks every call against the person's permissions, whatever the list shows.

Does TestingOps see our ChatGPT conversations?

No. The connection runs from ChatGPT to TestingOps: TestingOps receives the tool calls ChatGPT makes and their inputs, not the rest of the conversation. TestingOps agents calling outside tools is a different setup, not this one.

Next step

Request a demo and ask ChatGPT about the release readiness of one of your own projects.

Request a demo More connection guides integrations