APIs for AI agent function calling

Whether you use MCP or have custom code, Proxy optimizes your AI Agent to be faster, cheaper, and more accurate at calling tools.

Get Started

from openai import OpenAI

# Initialize LLM and Proxy
openai = OpenAI(api_key="openai_key_...")

# Pass functions to your agent to use.
completion = client.chat.completions.create(
  model="gpt-4o",
  messages=[{"role": "user", "content": user_prompt}],
  tools=tools,
)

Built for Production

Everything you need to deploy function calling at scale.

Optimized for Accuracy

Proxy prioritizes relevant context, achieving over 90% accuracy — surpassing models that rely on large, unfocused context windows.

~35%

Native LLMs

>90%

LLMs w/ Proxy

Secure Function Execution

Your functions are always executed in your environment - never on our servers - ensuring your code remains secure.

Your Environment

Schemas

Results

Proxy AI

Integrates Seamlessly

Our schema format works with any existing LLM you're using. Nothing needs to change about your current agent implementation.

Lower Cost, Higher Accuracy

Compared to improving function calling in-house, Proxy wins every time.

In-House

Requires researching the latest techniques

Fixed infrastructure costs

Manual conversion of code to schemas

Ongoing maintenance

Multiple services and providers

Ongoing, state-of-the-art upgrades

Pay-as-you-go pricing

Schemas are generated automatically

Fully managed service

One API, one provider

Join the Waitlist

Be the first to build AI agents at a new standard of performance.

Get Early Access