Claude
Overview
Claude is Anthropic's family of large language models known for extended context windows, strong reasoning capabilities, and careful instruction-following. Emergent natively supports Claude models through the Universal LLM Key, allowing you to use Claude for code generation, chat interactions, and other AI-powered features throughout the platform.
You can use Claude models in two ways:
- Automatically via the Universal LLM Key - Emergent provides access without requiring your own API key
- With your own API key - Connect your Anthropic account for direct billing and usage tracking
Using Claude with the Universal LLM Key
The simplest way to get started is using Emergent's Universal LLM Key, which provides immediate access to Claude models without any configuration.
The Universal LLM Key is included with your Emergent account and supports multiple AI providers including Anthropic. Your usage is metered as part of your Emergent subscription.
When you describe your app in chat, Emergent agents automatically use Claude models (or other supported providers) through the Universal LLM Key. No additional setup is required - just start building.
Connecting Your Own Claude API Key
The preferred approach is the Universal LLM Key. However, if you want to bring your own Claude API key, you can provide your own key when Emergent asks for your Claude credentials.
Visit console.anthropic.com and sign in to your Anthropic account. Create an API key and copy it.
Warning
Never commit your API key to version control or share it publicly. Treat it like a password.
Troubleshooting
| Issue | Solution |
|---|---|
| Claude responses are slower than expected | Larger models (Opus, Sonnet) prioritize quality over speed. If you need faster iteration, try Claude Haiku 4.5 for simpler tasks, then switch back to Sonnet for final polish. |
| Model returns 'overloaded' errors | Anthropic's API occasionally experiences high demand. Emergent will automatically retry. You can also manually retry the message or switch to another model temporarily. |
| Unexpected refusals or overly cautious responses | Claude is trained to decline potentially unsafe requests. If you believe a refusal is incorrect, rephrase your prompt to clarify intent, or try a different model for comparison. |
