keybridge integrations¶

keybridge is an OpenAI-compatible HTTP proxy — any framework or client that sends standard HTTP requests to LLM providers can use it without code changes beyond base_url.

Supported frameworks¶

Framework	Guide	Notes
CrewAI	crewai.md	Via `langchain_openai.ChatOpenAI` or `langchain_anthropic.ChatAnthropic`
LangGraph / LangChain	langgraph.md	Drop-in `base_url` swap; tools and streaming unchanged
AutoGen	autogen.md	Via `config_list` + `base_url`
OpenAI Agents SDK	autogen.md	Via custom `AsyncOpenAI` client
antcrew-platform	proxy index	Native integration with token management UI

Any OpenAI-compatible client¶

If your framework uses any OpenAI-compatible client, set base_url to keybridge's path for your provider:

import openai

client = openai.OpenAI(
    base_url="https://keybridge.yourcompany.com/openai",
    api_key=os.environ["KEYBRIDGE_TOKEN"],
)

For Anthropic via the Anthropic SDK:

import anthropic

client = anthropic.Anthropic(
    base_url="https://keybridge.yourcompany.com/anthropic",
    api_key=os.environ["KEYBRIDGE_TOKEN"],
)

How the token works¶

You generate a UUID token and give it to keybridge as PROXY_TOKEN. Your application uses that token as the api_key in all requests. keybridge validates the token, strips it, and injects the real provider key before forwarding upstream.

The real provider keys never leave your keybridge container.

your app  →  POST /openai/v1/chat/completions  →  keybridge
                 Authorization: Bearer <uuid-token>    ↓ validates token
                                                  strips it, injects OPENAI_API_KEY
                                                       ↓
                                                  api.openai.com

Generating a token without a platform¶

If you're not using a managed platform:

python -c "import uuid; print(uuid.uuid4())"

Use the output as both PROXY_TOKEN (keybridge env var) and KEYBRIDGE_TOKEN (application env var).