TestingOps.ai

AI model providers · Model provider

Anthropic Claude and TestingOps: test agents on your own Claude key

Save an Anthropic API key in TestingOps and the agents working on your projects run Claude through your own Anthropic account, on the model you pick.

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

In short

Anthropic Claude and TestingOps in short

The Anthropic connection is a bring-your-own-key setting: your customer admin saves an Anthropic API key under Integrations → AI models, and TestingOps stores it encrypted and attaches it to every project of your organisation.

While Anthropic is first in your provider list, the agents working on your projects run on the Claude adapter with that key, so their model usage is billed to your Anthropic account.

Without a key of your own, Claude runs on your projects start only if your package includes an AI allowance or use of the platform key has been allowed for your account; otherwise they stop before any request reaches a model.

What it does

What it does for your testing

  1. 1

    Agent work billed to your Anthropic account

    Input
    Your Anthropic API key, first under Your providers
    TestingOps
    Moves your organisation's agents to the Claude adapter and attaches the key to all of your projects
    Output
    Agent runs on your projects that use your key, billed by Anthropic to you
  2. 2

    One Claude model for every agent, or each agent's own

    Input
    A model name in the Model field, or none
    TestingOps
    With a model, moves your agents to it at their next run; without one, agents already on Claude keep their model
    Output
    A model choice you can predict across your projects
  3. 3

    Key problems found before work starts

    Input
    A revoked key, a wrong model name or terms not yet accepted
    TestingOps
    Test connection sends one short request with the key and model; before a run on your key, TestingOps also checks the key against Anthropic and keeps the result for five minutes
    Output
    Anthropic's own error message on the task instead of a run that fails halfway
  4. 4

    The same key on RunningOps

    Input
    An organisation linked to RunningOps
    TestingOps
    Copies the key and your provider order to the linked RunningOps organisation as TestingOps · Anthropic (Claude)
    Output
    Execution-side model settings without typing the key twice

Step by step

How it works

  1. Add Anthropic as a providerYour customer admin opens Integrations → AI models and, in the LLM card, picks Anthropic (Claude) under Add a model provider, then pastes the API key. The Model field is optional.
  2. The key is stored and attachedTestingOps keeps the key encrypted and attaches it to every project of your organisation, including projects added later. The value is never shown again; you replace it by typing a new one.
  3. The first provider drives the agentsUnder Your providers, the first entry is marked Default · drives agents. Saving a key or the order moves your organisation's agents to that provider's adapter and model from their next run; a run in progress keeps its adapter.
  4. Each run checks the key firstBefore a run uses your key, TestingOps checks it against Anthropic. If Anthropic rejects it, the run does not start and the task shows Anthropic's message.
  5. An empty balance moves runs alongIf Anthropic reports that your credit balance has run out, TestingOps holds that key for a while and sends runs to the next provider in your list, if you have one.

This setting decides which models your agents run on. Using TestingOps from claude.ai, Claude Desktop or Claude Code is the opposite direction and has its own connection guide.

Connect

Setup

  1. In the Claude Console, create an API key on the API keys page.
  2. In TestingOps, as customer admin, open Integrations → AI models and pick Anthropic (Claude) under Add a model provider in the LLM card.
  3. Paste the key into API key, enter a Model if all agents should use one model, and select Save.
  4. Open the entry again and select Test connection: TestingOps sends one short request with the saved key and model.
  5. If Anthropic (Claude) is not marked Default · drives agents, move it to the top under Your providers and select Save order.

Documentation: Claude API overview

Where you connect it in TestingOps: Integrations → AI models → Your providers → Anthropic Claude

Use cases

Use cases

Keep test work on your own Anthropic contract

Trigger
Your company already buys Claude usage from Anthropic and wants test agents billed there
Uses
Your Anthropic API key, first under Your providers
Agent activity
Agents working on your projects run on the Claude adapter with your key
Result
Agent usage on your Anthropic bill instead of the platform's
Condition
Some TestingOps AI features outside agent runs use the platform's models; see Limits

Move every agent to one Claude model

Trigger
Agents on your projects run on different Claude models and results are hard to compare
Uses
The Model field of the Anthropic entry
Agent activity
Each of your agents moves to that model at its next run
Result
Runs on your projects that use the same model
Condition
Leave Model empty if agents should keep the Claude model set on them

Switch from another provider to Claude

Trigger
Your agents run on another provider and you want them on Claude
Uses
The order under Your providers
Agent activity
Moving Anthropic (Claude) to the top and selecting Save order switches your agents to the Claude adapter
Result
Each agent's next run uses Claude; runs already in progress finish on their current adapter

Rotate a key without confusing failures

Trigger
Your security team rotates the Anthropic key
Uses
The API key field and Test connection
Agent activity
A run started with the revoked key stops before it begins, with Anthropic's message on the task
Result
The admin pastes the new key, tests it and the work continues
Condition
Only the customer admin, or a company admin for your organisation, can replace the key

Capabilities

Capabilities and limits

What is read

  • The list of models Anthropic offers your key, for the Model suggestions and the check before a run
  • Claude's responses while agents work on your projects

What is produced

  • Agent runs on your projects that use Claude with your key
  • A Reachable or Unreachable mark on the provider entry after Test connection

Where it can write

Nothing in your Anthropic account. TestingOps only sends model requests with your key, and Anthropic bills them to you.

How it connects

Anthropic's API with your API key; agents run on the Claude adapter. When your organisation is linked to RunningOps, the key and the provider order are copied there.

Account and permissions

  • An Anthropic API key, created on the API keys page of the Claude Console
  • Customer admin rights in TestingOps, or a company admin of the TestingOps workspace who manages keys for your organisation

Limits

  • One key per organisation, used on all of its projects: there is no per-project key
  • Agents follow the first entry under Your providers, so Anthropic has to be first for them to switch to Claude
  • Not every TestingOps AI feature uses your key: generating cases from a prompt, for example, runs on the platform's models
  • Without your key, Claude runs need a package AI allowance or permission to use the platform key; when the allowance is used up they stop with an AI usage limit message

Example

Example: how a team uses it

Illustrative scenario, not a customer story.

A QA team's company has an Anthropic agreement. The customer admin adds the company key under AI models, enters the Claude model the team has agreed on and moves Anthropic (Claude) to the top of Your providers. From their next run, the agents writing and reviewing the team's test cases use Claude with that key.

Weeks later the key is rotated. The next run stops before it starts and the task shows Anthropic's authentication error, so the admin pastes the new key, selects Test connection and the agents carry on.

Frequently asked questions

Is this the same as connecting Claude to TestingOps?

No. This setting decides which models TestingOps agents run on. Using TestingOps from claude.ai, Claude Desktop or Claude Code is the opposite direction: Claude calls TestingOps' MCP server, which you set up under Integrations → API & MCP.

What happens if we do not add an Anthropic key?

Agents on your projects can then use the platform's models, but only if your package includes an AI allowance (within its limit) or use of the platform key has been allowed for your account. Without either, a Claude run stops before it sends any request, with a message that names the missing key and points to Integrations → AI models. The note on the AI models screen that runs will not start reflects only the second permission, so check your package's allowance too.

Does all of our project data then go through our Anthropic account?

Agent runs on your projects do, while Anthropic is your first provider. We do not claim more than that: some TestingOps AI features, such as generating cases from a prompt, use the platform's models.

Which Claude model do the agents use?

The model you enter in Model, for all of your organisation's agents. If you leave it empty, agents already running on Claude keep their own model and agents that switch to Claude get a default Claude model. The suggestions in the field come from Anthropic's model list for your key.

Can two projects use different Anthropic keys?

No. The key belongs to your organisation and is attached to all of its projects, including new ones.

Who can see or change the key?

After saving, nobody can read it back: it is stored encrypted and the screen only shows that a key exists. The customer admin, or a company admin acting for your organisation, can replace or remove it.

Next step

Request a demo to see the AI models settings and an agent run on your own Anthropic key.

Request a demoMore ai models integrations