claude
Anthropic Claude models via inference.sh CLI. Models: Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5, plus Claude Code variants that run on your own Claude Pro/Max plan. 1M context, 128k output, thinking with effort, vision, tool use, structured output. Use for: coding, agents, reasoning, long documents, summarization, image understanding, chat. Triggers: claude, anthropic, claude api, claude fable, claude opus, claude sonnet, claude haiku, opus 5.5, sonnet 5.5, haiku 5.5, fable 5.1, claude code, claude pro, claude max, setup-token, anthropic api, claude subscription
Try your first task
A source-grounded starter task has not been verified for this skill yet.
SKILL.md
name: claudedescription: "Anthropic Claude models via inference.sh CLI. Models: Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5, plus Claude Code variants that run on your own Claude Pro/Max plan. 1M context, 128k output, thinking with effort, vision, tool use, structured output. Use for: coding, agents, reasoning, long documents, summarization, image understanding, chat. Triggers: claude, anthropic, claude api, claude fable, claude opus, claude sonnet, claude haiku, opus 5.5, sonnet 5.5, haiku 5.5, fable 5.1, claude code, claude pro, claude max, setup-token, anthropic api, claude subscription"allowed-tools: Bash(belt *)Install the belt CLI skill:
npx skills add belt-sh/cli
Claude
Run Anthropic's Claude models via inference.sh CLI.
Quick Start
Requires inference.sh CLI (
belt). Install instructions
belt login
belt app run anthropic/claude-sonnet-5-5 --input '{"text": "Explain how a B-tree works"}'
Models
| Model | App ID | Best For |
|---|---|---|
| Claude Fable 5.1 | anthropic/claude-fable-5-1 | Most capable: demanding reasoning, long-horizon agents |
| Claude Opus 5.5 | anthropic/claude-opus-5-5 | Long-running agentic coding and knowledge work |
| Claude Sonnet 5.5 | anthropic/claude-sonnet-5-5 | Fast frontier model for coding, agents, everyday work |
| Claude Haiku 5.5 | anthropic/claude-haiku-5-5 | Fastest, lowest cost, high-volume and latency-sensitive work |
All four: 1M context, 128k output, vision, tool use, direct Anthropic API. Fable 5.1 has always-on thinking with effort; Opus and Sonnet have adaptive thinking with effort; Haiku has adaptive thinking.
Claude Code (your own plan)
| Model | App ID |
|---|---|
| Claude Fable 5.1 (Claude Code) | anthropic/claude-code-fable-5-1 |
| Claude Opus 5.5 (Claude Code) | anthropic/claude-code-opus-5-5 |
| Claude Sonnet 5.5 (Claude Code) | anthropic/claude-code-sonnet-5-5 |
| Claude Haiku 5.5 (Claude Code) | anthropic/claude-code-haiku-5-5 |
These run the same models through Claude Code on your own Claude Pro or Max plan. Usage counts against your Claude plan, not inference.sh credits. Same inputs as the direct API apps.
Older versions (Opus 5, Opus 4.x, Sonnet 5, Haiku 4.5 and their Claude Code variants) are still available: belt app search claude.
Inputs
| Field | Description |
|---|---|
text | The user message |
system_prompt | System instructions |
images | Image files or URLs (vision) |
files | Document files |
context | Prior messages: [{"role": "user" | "assistant", "text": "..."}] |
reasoning_effort | none (default), low, medium, high |
temperature | Default 0.7 |
max_tokens | Default 64000 |
tools | Function tool definitions |
tool_choice | {"mode": "auto" | "none" | "required" | "function", "name": ...} |
response_format | {"type": "text" | "json_object" | "json_schema", ...} |
Output: response, reasoning, tool_calls, usage.
Each app also has an openai function (OpenAI Chat Completions compatible, takes messages): belt app sample anthropic/claude-sonnet-5-5 -f openai.
Examples
Coding with Opus
belt app run anthropic/claude-opus-5-5 --input '{
"system_prompt": "You are a senior Go engineer. Return only code.",
"text": "Write a rate limiter using a token bucket with tests",
"reasoning_effort": "high"
}'
Hard Reasoning with Fable
belt app run anthropic/claude-fable-5-1 --input '{
"text": "Design a migration plan to split a monolith Postgres schema into three services with zero downtime",
"reasoning_effort": "high"
}'
High Volume with Haiku
belt app run anthropic/claude-haiku-5-5 --input '{
"text": "Classify the sentiment of this review as positive, negative or neutral: The battery died after two days.",
"max_tokens": 10
}'
Vision
belt app run anthropic/claude-sonnet-5-5 --input '{
"text": "What is wrong in this UI screenshot?",
"images": ["https://example.com/screenshot.png"]
}'
Multi-turn Conversation
belt app run anthropic/claude-sonnet-5-5 --input '{
"context": [
{"role": "user", "text": "My name is Ada and I write Rust."},
{"role": "assistant", "text": "Nice to meet you, Ada."}
],
"text": "Suggest a weekend project for me."
}'
Structured Output
belt app run anthropic/claude-sonnet-5-5 --input '{
"text": "Extract the company and amount: Acme Corp raised $12M in Series A.",
"response_format": {
"type": "json_schema",
"name": "funding",
"json_schema": {
"type": "object",
"properties": {"company": {"type": "string"}, "amount": {"type": "string"}},
"required": ["company", "amount"]
}
}
}'
Tool Use
belt app run anthropic/claude-opus-5-5 --input '{
"text": "What is the weather in Paris?",
"tools": [{
"type": "function",
"function": {
"name": "get_weather",
"description": "Get current weather for a city",
"parameters": {"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"]}
}
}],
"tool_choice": {"mode": "auto"}
}'
The call comes back in tool_calls. Send the result as a context message with "role": "tool" and tool_call_id.
Claude Code Setup
- Install Claude Code and log in with your Claude account.
- Run
claude setup-tokenand copy the token (starts withsk-ant-oat01-, valid one year). - Save it on inference.sh:
belt credentials connect claude-code
- Run any Claude Code app:
belt app run anthropic/claude-code-opus-5-5 --input '{"text": "Refactor this function to be iterative: <code>"}'
The token is used only for your own runs, never shared with your workspace. Check how much of your plan is used (5-hour and weekly windows, with reset times):
belt app run anthropic/claude-code-opus-5-5 --function usage --input '{}'
usage has no inference.sh charge; it sends one token as a quota check against your plan.
Pricing
| Model | Input | Output | Cache read | Cache write |
|---|---|---|---|---|
| Fable 5.1 | $10.00/M | $50.00/M | $0.25/M | $12.50/M |
| Opus 5.5 | $4.00/M | $20.00/M | $0.20/M | $5.00/M |
| Sonnet 5.5 | $2.00/M | $10.00/M | $0.20/M | $2.50/M |
| Haiku 5.5 (<=100K tokens) | $0.10/M | $0.50/M | - | - |
| Haiku 5.5 (>100K tokens) | $0.50/M | $2.50/M | - | - |
Claude Code apps bill your Claude plan. Estimate a run: belt app estimate <app-id> --input input.json.
Related Skills
# Full platform skill (all apps)
npx skills add inference-sh/skills@infsh-cli
# All LLMs in one place
npx skills add inference-sh/skills@llm-models
# Other providers
npx skills add inference-sh/skills@gpt
npx skills add inference-sh/skills@gemini
# Agents and tool use
npx skills add inference-sh/skills@agent-tools
# Search + summarize papers with Claude
npx skills add inference-sh/skills@academic-research
Browse all apps: belt app list
Documentation
- Claude Code Subscription - Using your own Claude plan
- Agents Overview - Building AI agents