ironclaw/docs/drafts/providers/moonshot.mdx

---
title: Moonshot AI
description: Kimi K2.5 and other Moonshot models via OpenAI-compatible API
---

[Moonshot AI](https://platform.moonshot.ai) provides state-of-the-art large language models including Kimi K2.5, featuring advanced reasoning capabilities and an extensive context window.

## Overview

Moonshot AI offers the Kimi family of models through an OpenAI-compatible API:

- **Kimi K2.5** — State-of-the-art reasoning with 256K context window
- **Kimi K1.6** — Long-context model for document analysis
- **OpenAI-compatible** — Standard `/v1/chat/completions` endpoint
- **Competitive pricing** — Pay-as-you-go token-based billing

## Configuration

```bash
# ~/.ironclaw/.env

LLM_BACKEND=openai_compatible
LLM_BASE_URL=https://api.moonshot.ai/v1
LLM_API_KEY=sk-...
LLM_MODEL=kimi-k2-5
```

## Getting an API Key

1. Visit [platform.moonshot.ai](https://platform.moonshot.ai)
2. Create an account and complete verification
3. Generate an API key from the dashboard
4. Copy the key and add it to your `.env` file

## Available Models

| Model | Context Window | Description |
|-------|----------------|-------------|
| `kimi-k2-5` | 256,000 tokens | Flagship reasoning model (recommended) |
| `kimi-k1.6` | 2,000,000 tokens | Ultra-long context for documents |
| `kimi-k2-5-instruct` | 256,000 tokens | Instruction-tuned variant |

<Note>
Model IDs may vary. Check the [Moonshot documentation](https://platform.moonshot.ai/docs/overview) for the latest available models and exact IDs.
</Note>

## Setup

### Via Wizard

```bash
ironclaw onboard
```

1. Step 3: Select "Other (OpenAI-compatible)"
2. Enter base URL: `https://api.moonshot.ai/v1`
3. Enter your API key
4. Step 4: Enter model name: `kimi-k2-5`

### Manual Configuration

```bash
# Add to ~/.ironclaw/.env
export LLM_BACKEND=openai_compatible
export LLM_BASE_URL=https://api.moonshot.ai/v1
export LLM_API_KEY=sk-your-moonshot-key
export LLM_MODEL=kimi-k2-5
```

Restart IronClaw:
```bash
ironclaw run
```

## Using Kimi K2.5

Kimi K2.5 is Moonshot's flagship model with advanced capabilities:

### Strengths

- **Long context** — 256K token window for large documents
- **Strong reasoning** — Excellent for complex analysis
- **Multilingual** — Strong performance in Chinese and English
- **Tool use** — Supports function calling

### Example Use Cases

<CodeGroup>
```bash Document Analysis
# Kimi's long context excels at analyzing large documents
"Summarize this 100-page PDF and extract key findings"
```

```bash Code Review
# Strong reasoning for code analysis
"Review this Rust module for potential bugs and improvements"
```

```bash Complex Reasoning
# Multi-step problem solving
"Analyze these three approaches and recommend the best one with justification"
```
</CodeGroup>

## API Compatibility

Moonshot implements the OpenAI Chat Completions API:

```
POST https://api.moonshot.ai/v1/chat/completions
```

Supported features:
- Streaming responses (`stream: true`)
- Function calling / tool use
- System messages
- Temperature and top-p sampling

<Note>
Some OpenAI-specific features like logprobs may not be available. Refer to Moonshot's API documentation for the latest feature support.
</Note>

## Pricing

Moonshot uses pay-as-you-go pricing based on tokens consumed:

| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|-------|----------------------|------------------------|
| Kimi K2.5 | Check current rates | Check current rates |
| Kimi K1.6 | Check current rates | Check current rates |

Visit [platform.moonshot.ai/pricing](https://platform.moonshot.ai/pricing) for current rates.

## Troubleshooting

<AccordionGroup>
  <Accordion title="API key invalid" icon="key">
    - Verify your API key is correctly copied
    - Ensure your account is verified
    - Check for any account restrictions
  </Accordion>

  <Accordion title="Model not found" icon="search">
    - Confirm the exact model ID from the Moonshot dashboard
    - Try `kimi-k2-5` or `kimi-k2-5-instruct`
    - Check the model is available in your region
  </Accordion>

  <Accordion title="Rate limiting" icon="gauge">
    - Moonshot may rate-limit requests
    - Implement exponential backoff for retries
    - Consider upgrading your plan for higher limits
  </Accordion>

  <Accordion title="Context length exceeded" icon="maximize">
    - Kimi K2.5 supports 256K tokens
    - Use `memory_write` to store large documents
    - Chunk large inputs when possible
  </Accordion>
</AccordionGroup>

## Comparison with Tinfoil

Both providers offer Kimi K2.5, but with different tradeoffs:

| Feature | Moonshot | Tinfoil |
|---------|----------|---------|
| **Model** | Kimi K2.5 | Kimi K2.5 |
| **Privacy** | Standard | TEE-encrypted |
| **Pricing** | Direct pay-as-you-go | Tinfoil rates |
| **Attestation** | No | Hardware-verified |
| **Use case** | General use | Maximum privacy |

Choose **Moonshot** for direct API access and competitive pricing. Choose **Tinfoil** when you need TEE privacy guarantees.

## Next Steps

<CardGroup cols={2}>
  <Card title="Tinfoil" icon="shield" href="/providers/tinfoil">
    TEE-secured Kimi inference alternative
  </Card>

  <Card title="OpenAI-Compatible" icon="layers" href="/providers/openai-compatible">
    Other compatible providers
  </Card>

  <Card title="Configuration" icon="settings" href="/setup/configuration">
    Full environment variable reference
  </Card>

  <Card title="OpenAI" icon="circle" href="/providers/openai">
    Compare with GPT models
  </Card>
</CardGroup>