Optimization Layer for AI Agents

Regardless of how you build your AI agent, Proxy makes it faster, cheaper, and more accurate when using tools.

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,
)

How Proxy Works

Cut your token costs and latency in just 3 steps.

Index Your Functions

Provide the tools that you want your AI agent to have access to. Schemas for OpenAI, Anthropic, and MCP tools are supported.

Your Environment

Tools

Results

Proxy AI

Send User Prompts

Forward prompts to our search endpoint and we'll return back the best functions for the task at hand. Set any number you'd like returned.

Cut Token Costs

Save thousands on input token costs annually by only including the tools your agent needs for each task.

100,000 Tokens

GPT-4o agent w/ 100 tools
(All 100 provided)

5,000 tokens

Same agent w/ Proxy
(Best 5 tools provided)

Simple, Transparent Pricing

Pay-as-you-go pricing that scales with you.

Upload

$0.36 / 100K tool tokens

Store your tools once and save on token costs forever.

Anthropic, OpenAI, and MCP formats supported.

Upload Python code to convert functions to tools.

Delete tools at any time for no cost.

$0.36 / 100K prompt tokens + $0.04 / 1K tools stored

Configure the amount of tools you want returned.

Automatic updates that continually improve results.

Filter results by tool format, file name, and more.

Boost your agent's accuracy for every type of task.

Demo Proxy Today

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

Book a Demo