AI AGENTS API
Build autonomous AI agents using function calling and tool use through Runcrate's OpenAI-compatible API. DeepSeek, Llama, Qwen, Claude, and Mistral all support the standard tools format. Define functions, let the model decide when to call them, execute the results, and loop. Build agents that browse the web, query databases, call APIs, and complete multi-step tasks.
QUICK START
from openai import OpenAI
client = OpenAI(
base_url="https://api.runcrate.ai/v1",
api_key="rc_live_YOUR_API_KEY",
)
tools = [
{
"type": "function",
"function": {
"name": "get_weather",
"description": "Get current weather for a city",
"parameters": {
"type": "object",
"properties": {
"city": {"type": "string", "description": "City name"},
},
"required": ["city"],
},
},
}
]
response = client.chat.completions.create(
model="deepseek-ai/DeepSeek-V3",
messages=[{"role": "user", "content": "What is the weather in Tokyo?"}],
tools=tools,
)
# The model returns a tool_call you can execute
tool_call = response.choices[0].message.tool_calls[0]
print(tool_call.function.name, tool_call.function.arguments)AVAILABLE MODELS
| Model | Provider | Price | Detail |
|---|---|---|---|
| deepseek-ai/DeepSeek-V3 | DeepSeek | Per-token | Strong tool use, 128K context |
| anthropic/claude-4-sonnet | Anthropic | Per-token | Reliable function calling |
| Qwen/Qwen3-32B | Alibaba | Per-token | 32B, strong tool use |
| meta-llama/Llama-4-Scout-17B-16E-Instruct | Meta | Per-token | MoE, efficient agent loops |
WHY RUNCRATE
Standard OpenAI function calling format. Define tools with JSON Schema, the model decides when and how to call them. Works with LangChain, CrewAI, and every OpenAI-compatible agent framework.
Models can request multiple tool calls in a single response. Execute them concurrently for faster agent loops and reduced round trips.
Use different models for different agent tasks. DeepSeek for reasoning, Claude for analysis, Llama for fast tool dispatch. One API key covers all.
Build agents that connect to MCP servers for tool discovery and execution. Combine model intelligence with external capabilities.
FAQ