BlockRun AI
Access 78 AI models and image generation via BlockRun. Use a BlockRun account API key, or pay through the lobster.cash Solana wallet via x402.
Account mode uses BlockRun credits. Wallet mode delegates Solana payment execution to lobster.cash.
When to Activate
Use this skill when the user:
- Wants to chat with an AI model (GPT, Claude, Gemini, DeepSeek, etc.)
- Wants to generate images (DALL-E 3, GPT Image 1, Flux)
- Asks about available AI models or pricing
- Mentions BlockRun or pay-per-request AI
Tools Overview
| Tool | Purpose |
|---|---|
blockrun_models |
List available models and pricing |
blockrun_chat |
Chat with any AI model |
blockrun_image |
Generate images with AI |
Prerequisites
For account access, direct the user to register, create a key, and add credits. The user sets BLOCKRUN_API_KEY in the gateway environment and restarts it. Never ask them to paste credentials into chat or tool arguments. All three tools use that key without reading wallet files.
Without a key, set up the lobster.cash Solana wallet and fund it with USDC. Base access is available through the main BlockRun SDK; this adapter only implements Solana wallet signing.
Common Operations
List available models
User: "What AI models are available?"
Agent: Use blockrun_models
User: "Show me OpenAI models"
Agent: Use blockrun_models with filter="openai"
Chat with a model
User: "Ask GPT-5 what the capital of France is"
Agent: Use blockrun_chat with model="openai/gpt-5.2", message="What is the capital of France?"
User: "Use DeepSeek to explain quantum computing"
Agent: Use blockrun_chat with model="deepseek/deepseek-chat", message="Explain quantum computing simply"
Generate images
User: "Generate an image of a sunset over the ocean"
Agent: Use blockrun_image with prompt="A beautiful sunset over the ocean with vibrant orange and purple colors"
Available Models
Chat Models (selection)
| Provider | Models | Pricing (per 1M tokens) |
|---|---|---|
| OpenAI | gpt-5.2, gpt-5-mini, gpt-4.1 | $1.75–$14/M output |
| Anthropic | claude-opus-4.6, claude-sonnet-4.6, claude-haiku-4.5 | $1–$15/M output |
| gemini-2.5-pro, gemini-2.5-flash | $0.30–$10/M output | |
| DeepSeek | deepseek-chat | $0.28/M output |
| xAI | grok-4-1-fast-reasoning | varies |
| NVIDIA | gpt-oss-120b | FREE |
Use blockrun_models for the full up-to-date list.
Image Models
| Model | Price per image |
|---|---|
| openai/dall-e-3 | ~$0.04 |
| openai/gpt-image-1 | ~$0.02 |
| black-forest/flux-1.1-pro | ~$0.04 |
How Payment Works
With BLOCKRUN_API_KEY, requests go to https://api.blockrun.ai/v1 and charge account credits. A 401, 402 or 429 is an account error; do not switch to a wallet or automatically replay a paid call.
Without a key, BlockRun uses x402. Transaction execution and final status are handled by lobster.cash:
- Tool sends the AI request to
sol.blockrun.ai - If payment is required, lobster.cash handles wallet signing and settlement
- AI response is returned
Choose account credits or Solana wallet billing.
Tool Parameters
blockrun_models
{
"filter": "optional - keyword to filter (e.g., 'openai', 'free', 'image')"
}
blockrun_chat
{
"model": "required - model ID (e.g., 'openai/gpt-5.2')",
"message": "required - the message to send",
"systemPrompt": "optional - system prompt for model behavior",
"maxTokens": "optional - max response tokens (default: 1024)",
"temperature": "optional - creativity 0-2 (default: 1)"
}
blockrun_image
{
"prompt": "required - description of the image to generate",
"model": "optional - 'openai/dall-e-3' (default), 'openai/gpt-image-1', or 'black-forest/flux-1.1-pro'",
"size": "optional - '1024x1024' (default), '1792x1024', or '1024x1792'",
"quality": "optional - 'standard' (default) or 'hd'"
}
Troubleshooting
- Model not found — Use
blockrun_modelsto verify the model ID - Invalid key / insufficient credits — use the account key or credits pages above; do not request a wallet deposit in account mode
- Payment failed — Check wallet USDC balance via lobster.cash wallet tools
- Empty response — Try a different model or adjust max_tokens
Best Practices
- Check models first — Use
blockrun_modelsto see what's available and pricing - Use cheaper models for simple tasks — DeepSeek or free NVIDIA models for basic queries
- Be specific with image prompts — Detailed prompts produce better results
- Set appropriate max_tokens — Lower values save costs for short responses