Claude models in Taskade run on your own Anthropic key. You can pick Claude on an agent with Bring Your Own Key, or add Claude as a step in an automation. Both paths bill your Anthropic account, not your Taskade AI credits.
Claude models favor finish quality. They produce a polished result, and they handle follow-up changes well.
Claude models you can run on your own key
| Claude model | Good at |
|---|---|
| Claude Opus 5.5 | The newest Opus model, with extended thinking for the hardest tasks |
| Claude Opus 5 | Maximum reasoning with extended thinking |
| Claude Sonnet 5.5 | The newest Sonnet model, with extended thinking for complex tasks |
| Claude Sonnet 5 | Extended thinking for complex tasks |
| Claude Haiku 4.5 | Fast, efficient everyday work |
How to pick within Claude
- Opus 5.5 or Opus 5 when the result is the deliverable. Opus holds a complete workspace in view.
- Sonnet 5.5 or Sonnet 5 for most real work. Sonnet is the balance point of the family, and it revises its own output well.
- Haiku 4.5 for volume. Use it where the request is small and you want the answer fast.
A very long instruction can lose a detail near the end. Split a large build into steps, and send one step at a time.
Run Claude on an agent with your own key
Bring Your Own Key is available on Pro and up. It bills the agent's model usage to your Anthropic account, not to your Taskade plan.
- Open the agent's settings.
- Under AI Model, choose a Claude model.
- Under Bring Your Own Key, select your Anthropic key.
If Taskade finds no Anthropic key for the agent, the settings show "Claude runs only on your own Anthropic key. Without one, this agent runs on Auto."
Ask Claude in an automation
The Anthropic Claude connector runs Ask Claude as a step in any automation. Add the step, connect your Anthropic account, and feed it a prompt from an earlier step. The connector also carries Extract Structured Data and Ask About an Image.
The step runs on your key, so it draws no Taskade AI credits. See Multi-model AI access for the setup.
A saved Claude pick keeps working
A chat, agent, or automation step can keep a Claude model in its settings. If Taskade has no Anthropic key for it, Taskade runs each turn on TSK-Auto. The turn completes with no error and with no edit from you.
The rest of the catalog
With no key, Taskade offers frontier models from top AI labs: OpenAI GPT, xAI Grok, and open-weight models such as DeepSeek, Qwen, Kimi, and GLM. The full catalog lists the models that run on Taskade AI credits. You can also leave a surface on TSK-Auto or a TSK rung.
How Claude performs
We build the same apps on every family and publish what happened. See the Claude build results for the test-by-test record.
Next steps
- Compare the models in the full catalog.
- Read how Taskade picks a model for Auto and the rungs.
- See Gemini in Taskade to run Gemini on your own Google key in an automation.
- Open each Claude model from the Anthropic models page to see its plans and test results.