Skip to content

Enterprise deploymentCustomer-hosted or fully managed.Contact salesView pricing

← Back

+Integrations · 15

Works with what you already run.

Most integrations are one line: point the client at Intertrace and use a runtime key. Your code, prompts and models stay the same.

The whole change, for most apps:

Swap the base URL and use an itr_rt_* runtime key. Every model call, tool call and MCP request is then decided before it runs, and shows up in Traffic with the reason.

before → after
- base_url = "https://api.openai.com/v1"
- api_key  = OPENAI_API_KEY
+ base_url = "https://intertrace.fly.dev/v1"
+ api_key  = INTERTRACE_RUNTIME_API_KEY   # itr_rt_…

+SDKs & gateways

Swap the base URL; keep your code.

OpenAI SDK

▶ video

Chat Completions and Responses through Intertrace with one base URL change, in Python or Node.

Set up →

LiteLLM

Point LiteLLM's api_base at Intertrace and every provider you route through it is inspected.

Set up →

+Agent frameworks

Route every model call a framework makes.

LangChain

▶ video

ChatOpenAI against Intertrace — chains and agents keep working, every call is decided.

Set up →

LlamaIndex

RAG pipelines whose model calls and retrieved content pass through the layer.

Set up →

CrewAI

▶ video

Protect a whole crew by pointing its LLM at Intertrace — no agent rewrites.

Set up →

OpenAI Agents SDK

Set a default client once; every agent run goes through the layer.

Set up →

LangGraph

Graph agents on the Intertrace edge, with tools wrapped for pre-action authorization.

Set up →

Mastra

Use the runtime package for edge config and wrap tools with pre-action checks.

Set up →

Pydantic AI

An OpenAI-compatible model pointed at Intertrace.

Set up →

+Coding agents

Guardrails on the agents that write and run code.

Codex

A Codex provider profile that sends Responses traffic through Intertrace.

Set up →

Cursor

▶ video

Override the OpenAI base URL in Cursor and add MCP through the Intertrace edge.

Set up →

GitHub Copilot

Business and Enterprise proxy override to the Intertrace edge.

Set up →

+Chat apps

Put self-hosted chat UIs behind policy.

Claude Desktop

Remote MCP through Intertrace, with optional inference via the /anthropic route.

Set up →

Open WebUI

Run Open WebUI with Intertrace as its OpenAI endpoint.

Set up →

+MCP

Authorize every tool call before it runs.

MCP servers

▶ video

Put MCP tool calls in the request path for Cursor, Claude and Devin — authorized, pinned and logged.

Set up →

+Anything else

Speaks OpenAI? It works.

Any client that lets you set an OpenAI-compatible base URL can go through Intertrace. For raw HTTP, see the curl guide in the docs.

curl / raw HTTP →Models & providers →Quickstart →

+ The AI watching your AI

Connected in minutes.

Get a runtime key, change one line, and watch your first decision land in Traffic.

Book a demoTry the Simulation Lab