Assistants & MCP

OpenAI Agents & Operator MCP Integration

Connect the OpenAI Agents SDK and Operator workflows to OneInfer via the Model Context Protocol (MCP) — granting agents autonomous access to 100+ models, image generators, and GPU instances.

Provide tools to OpenAI Agents by pointing their tool provider to https://mcp.oneinfer.ai/mcp. No custom tool schema coding required — tool definitions and input schemas are served dynamically via MCP.

OpenAI Agent MCP Tool Architecture

The OpenAI Agents SDK discovers OneInfer tools dynamically, executing model completions, credit inquiries, and GPU machine queries directly over SSE.

Layer 1

OpenAI Agents SDK / Operator

Autonomous orchestration loop and task delegation.

Layer 2

OneInfer MCP Server

https://mcp.oneinfer.ai/mcp

Layer 3

Tool Operations

Chat completions, image/video jobs, GPU nodes, and storage.

Connecting OpenAI Agents to OneInfer MCP

Attach OneInfer MCP tools directly to an OpenAI Agent:

python

from openai_agents import Agent, Runner
from mcp_client import MCPToolProvider

# Initialize OneInfer MCP Tool Provider
mcp = MCPToolProvider(url="https://mcp.oneinfer.ai/mcp")

# Build agent with OneInfer MCP tools
cluster_agent = Agent(
    name="OneInfer Fleet Manager",
    instructions="Manage AI inference models and GPU machines using OneInfer MCP tools.",
    tools=await mcp.get_tools()
)

runner = Runner(agent=cluster_agent)
result = await runner.run("Check my available OneInfer credits and list active GPU nodes.")
print(result.final_output)