Anthropic (Claude)
Anthropic provides the Claude family of models, known for strong reasoning, safety alignment, extended thinking capabilities, and large context windows up to 1M tokens.
Getting an API Key
- Visit console.anthropic.com/settings/keys
- Sign in or create an account
- Generate a new API key (starts with
sk-ant-...) - Paste the key into AI Supreme Council under Settings > AI Model > Anthropic
Anthropic requires a credit card on file and prepaid credits before you can use the API. There is no free tier for direct API access. For free Claude access, use OpenRouter which offers some Claude models.
API keys are stored locally in your browser (localStorage) and are never included in shared bot URLs.
Supported Models
| Model | Context Window | Max Output | Input Price | Output Price | Capabilities |
|---|---|---|---|---|---|
| Claude Opus 4.6 | 1M | 128K | $5.00/MTok | $25.00/MTok | Vision, tools, code, streaming |
| Claude Sonnet 4.5 | 1M | 64K | $3.00/MTok | $15.00/MTok | Vision, tools, code, streaming |
| Claude Sonnet 4 | 200K | 64K | $3.00/MTok | $15.00/MTok | Vision, tools, code, streaming |
| Claude Opus 4 | 200K | 32K | $15.00/MTok | $75.00/MTok | Vision, tools, code, streaming |
| Claude Haiku 4.5 | 200K | 64K | $0.80/MTok | $4.00/MTok | Vision, tools, code, streaming |
Prices are per million tokens (MTok). All models support cached input at reduced rates.
- Claude Opus 4.6 is the flagship model -- best for complex reasoning, coding, and long-form analysis. Use it when quality matters most.
- Claude Sonnet 4.5 is the balanced choice -- strong performance at a lower price than Opus.
- Claude Haiku 4.5 is the speed and cost leader -- ideal for simple tasks, high-throughput workflows, and budget-conscious usage.
Extended Thinking
Claude models support extended thinking, where the model reasons through a problem step by step before delivering its final answer. Thinking output appears in a collapsible block above the response.
To enable extended thinking, set the Reasoning Effort in the bot configuration panel:
| Setting | Budget Tokens | Best For |
|---|---|---|
low | 8,192 | Quick reasoning, simple problems |
medium | 32,768 | Moderate analysis, standard tasks |
high | 128,000 | Deep reasoning, complex problems |
max | Model limit - 1024 | Maximum reasoning depth |
| Custom number | Your value | Fine-tuned control |
When extended thinking is enabled, Anthropic requires temperature=1 (the app handles this automatically). The max_tokens parameter is increased to accommodate both thinking and response tokens.
Vision Support
All Claude models support vision input. You can:
- Paste images directly into the chat input (Ctrl+V / Cmd+V)
- Upload images using the attachment button
- Drag and drop images into the chat area
Claude can analyze images, read text in screenshots, describe visual content, and reason about diagrams or charts.
Tool Calling
Claude supports function/tool calling, where the model can invoke structured functions you define. AI Supreme Council normalizes tool calls between Anthropic's format and the OpenAI format, so tools work consistently across providers.
Tool calls appear in the chat as formatted blocks showing the function name and arguments.
Pricing
Anthropic bills per input and output token. All models support prompt caching, which reduces input costs by up to 90% for repeated context (like system prompts). Refer to anthropic.com/pricing for the latest rates.
AI Supreme Council tracks token usage per provider in Settings > Usage, so you can monitor spending across sessions.
Configuration
When creating a bot profile, select Anthropic as the provider and choose your preferred model. You can override the global API key with a per-bot key in the bot configuration panel.
The Anthropic provider uses the Messages API (api.anthropic.com/v1/messages) with streaming enabled by default. The anthropic-version and anthropic-dangerous-direct-browser-access headers are set automatically to enable direct browser access.
Tips for Best Results
- Use system prompts effectively. Claude follows system prompt instructions carefully. Be specific about the format, tone, and constraints you want.
- Enable thinking for hard problems. Extended thinking significantly improves performance on math, logic, and multi-step reasoning tasks.
- Leverage the large context window. Claude Opus 4.6 and Sonnet 4.5 support 1M tokens of context -- paste in entire codebases, long documents, or detailed specifications.
- Use Haiku for iteration. When experimenting with prompts, use Haiku to test cheaply before switching to Opus or Sonnet for final results.