TestingOps.ai

AI model providers · Model provider · Connection guide

Google Gemini and TestingOps: agents on your Gemini key, TestingOps in Gemini CLI

Gemini connects to TestingOps in two independent ways: TestingOps agents can run on your own Gemini API key, and Gemini CLI can use TestingOps as a tool over MCP.

  • Model provider
  • Connection guide
  • Available
Request a demo Gemini API keys
Connection
Model provider, Connection guide
Writes back
Nothing, only model requests
Last checked
7 Nov 2026

In short

Google Gemini and TestingOps in short

This page covers two separate connections: TestingOps agents can run on your own Gemini API key, saved by the customer admin under Integrations → AI models, and Gemini CLI can use TestingOps over MCP with an access key.

With the first, the agents working on your projects run on the Gemini adapter and your key while Google Gemini is your first provider.

With the second, Gemini works as the person who created the access key: it can read runs, results and defects and, with read and write access, create scenarios and cases.

Each connection works without the other.

What it does

What it does for your testing

  1. 1

    Agents on your key: billed to your Google account

    Input
    A Gemini API key, first under Your providers
    TestingOps
    Moves your organisation's agents to the Gemini adapter and attaches the key to all of your projects
    Output
    Agent runs on your projects that use Gemini with your key
  2. 2

    Agents on your key: checked before a run

    Input
    Your saved Gemini key
    TestingOps
    Asks the Gemini API for its model list with the key before a run on it
    Output
    A rejected key stops the run before it starts, with Google's message on the task
  3. 3

    Over MCP: run results in your terminal

    Input
    Gemini CLI with the TestingOps MCP server configured
    TestingOps
    Lets Gemini call TestingOps tools as you: runs, results, failure analysis and defects
    Output
    Answers about your projects next to the code you are working on
  4. 4

    Over MCP: cases from a spec without leaving the terminal

    Input
    A requirement you give Gemini CLI and a Read and write access key
    TestingOps
    Gemini creates scenarios and cases in TestingOps through its MCP tools
    Output
    Cases stored in your TestingOps project, ready for review

Step by step

How it works

  1. Agents on your key: save a Gemini API keyThe customer admin picks Google Gemini under Add a model provider (Integrations → AI models) and saves the key. TestingOps stores it encrypted and attaches it to every project of your organisation as GEMINI_API_KEY and GOOGLE_API_KEY.
  2. Agents on your key: Gemini at the top drives the agentsWhile Google Gemini is marked Default · drives agents, your agents run on the Gemini adapter from their next run. Before a run, TestingOps checks the key against the Gemini API.
  3. Over MCP: create an access keyUnder Integrations → API & MCP, create a key under Access keys with Read only or Read and write access and an expiry of 30, 90 or 180 days or one year. The key is shown once.
  4. Over MCP: add TestingOps to Gemini CLIIn ~/.gemini/settings.json, add a server under mcpServers with httpUrl set to the Endpoint (streamable HTTP) shown on the MCP server card, and an Authorization header with Bearer and your key.
  5. Over MCP: Gemini works as youGemini sees the tools your account may use, in one company. A read-only key cannot change anything, and starting a run on a production environment needs explicit approval.

The Client config snippet on the MCP server card uses url; Gemini CLI expects httpUrl for a streamable HTTP server, with the same address.

Connect

Setup

  1. Agents on your key: create a Gemini API key in Google AI Studio.
  2. Agents on your key: as customer admin, open Integrations → AI models, pick Google Gemini under Add a model provider, paste the key into API key and select Save; then open the entry and select Test connection.
  3. Agents on your key: make sure Google Gemini is marked Default · drives agents under Your providers, or move it up and select Save order.
  4. Over MCP: open Integrations → API & MCP and copy the Endpoint (streamable HTTP) from the MCP server card.
  5. Over MCP: under Access keys enter a name, choose Read only or Read and write and an expiry, select Create key and copy the key, which is not shown again.
  6. Over MCP: in ~/.gemini/settings.json add a server under mcpServers with httpUrl set to the endpoint and headers containing Authorization with the value Bearer followed by the key.

Documentation: Gemini API keys · Gemini CLI MCP servers

Where you connect it in TestingOps: Integrations → AI models → Your providers → Google Gemini

Use cases

Use cases

Agents on your key: standardise test agents on Gemini

Trigger
Your company uses Gemini through Google AI Studio and wants test agents there too
Uses
Your Gemini API key, first under Your providers
Agent activity
Agents on your projects run on the Gemini adapter with your key
Result
Agent usage billed to your Google account
Condition
Some TestingOps AI features outside agent runs use the platform's models

Over MCP: ask about a failed run from the terminal

Trigger
A developer sees a red pipeline and wants the details without opening TestingOps
Uses
Gemini CLI with a Read only access key
Agent activity
Gemini lists the latest runs, reads the failed results and asks TestingOps for its failure analysis
Result
The failing steps and the likely cause, next to the code
Condition
The key reaches only what the developer's own account can see

Over MCP: turn a spec into cases from Gemini CLI

Trigger
A tester drafts acceptance criteria in the terminal
Uses
A Read and write access key
Agent activity
Gemini searches existing cases in TestingOps, then creates scenarios and cases for the new criteria
Result
New cases in the TestingOps project, created as the tester
Condition
New cases are reviewed like any others

Capabilities

Capabilities and limits

What is read

  • Agents on your key: the Gemini API's model list for your key, and Gemini's responses while agents work on your projects
  • Over MCP: what the key's owner may see in one company, such as test cases and scenarios, runs with results and artifacts, defects, knowledge-base search, traceability, release readiness and failure analysis; customer-portal accounts get a smaller tool set and only their own projects

What is produced

  • Agents on your key: agent runs on the Gemini adapter, billed to your Google account
  • Over MCP: answers in Gemini built from TestingOps data and, with Read and write access, new scenarios, cases, runs or defects in TestingOps

Where it can write

  • Agents on your key: nothing in your Google account; only model requests are sent
  • Over MCP: with Read and write access, Gemini can create scenarios and cases, start or cancel runs and report or verify defects in TestingOps; a read-only key is refused for every change

How it connects

  • Agents on your key: the Gemini API with your key, through the Gemini adapter
  • Over MCP: Gemini CLI calls the TestingOps MCP server (streamable HTTP, address ending in /api/mcp/v1) with an access key in the Authorization header

Account and permissions

  • Agents on your key: a Gemini API key from Google AI Studio, saved by the customer admin
  • Over MCP: a TestingOps access key created under Integrations → API & MCP; it acts as you, in one company, until it expires or you revoke it

Limits

  • Agents on your key: one key per organisation for all of its projects, and not every TestingOps AI feature uses it
  • Over MCP: TestingOps offers tools only, no MCP prompts or resources, and a key allows 600 requests per minute by default
  • Over MCP: signing in with /mcp auth instead of a key is not confirmed: Gemini CLI returns to an http://localhost address by default, and TestingOps accepts only https or numeric loopback (127.0.0.1) return addresses
  • Over MCP: Gemini Enterprise needs an administrator, sign-in and a TestingOps address reachable over public https

Example

Example: how a team uses it

Illustrative scenario, not a customer story.

A product team runs its TestingOps agents on Gemini: the customer admin saved the team's Gemini API key and moved Google Gemini to the top of Your providers, so agent runs bill to the team's Google account.

One of the developers also adds TestingOps to Gemini CLI with a read-only access key. After a failed nightly run she asks Gemini which cases failed and why; Gemini reads the run's results and failure analysis from TestingOps and points to the step that broke.

Frequently asked questions

Are the two Gemini connections related?

No. One decides which models TestingOps agents run on; the other lets Gemini call TestingOps. You can set up either one without the other, and the MCP connection does not need a Gemini API key.

What happens to Gemini runs without our own key?

A run on the Gemini adapter starts only if your package includes an AI allowance or use of the platform key has been allowed for your account. Otherwise it stops before any request, with a message that names the missing Gemini key.

Can Gemini change things in TestingOps?

Only with a Read and write key. It can then create scenarios and cases, start or cancel runs and report defects as you. A run on a production environment needs explicit approval, and a read-only key cannot change anything.

Why does our Gemini CLI configuration not connect?

Gemini CLI expects httpUrl for a streamable HTTP server, while the Client config snippet on the MCP server card uses url. Use httpUrl with the same address and keep the Authorization header.

Can we sign in from Gemini CLI instead of using a key?

Use a key for now. Gemini CLI's sign-in returns to an http://localhost address by default, and TestingOps accepts only https or numeric loopback (127.0.0.1) return addresses, so the sign-in path is not confirmed.

Does Gemini Enterprise work too?

TestingOps' help describes it: an administrator adds TestingOps as a custom MCP server data store with sign-in. It needs a TestingOps address reachable over public https, and this path has not been verified end to end.

Next step

Request a demo to see both Gemini connections: agents on your Gemini key, and TestingOps answering inside Gemini CLI.

Request a demoMore ai models integrations