Frameworks
Meter and attribute the spend of agents built with popular orchestration frameworks — with no rewrite.
Silvanexum meets your agents where they already live. If you build with an established orchestration framework, these guides show how to bring your agent's spend under control without changing your logic.
Every guide follows the same one-step pattern: point the framework's model client
at the Silvanexum gateway (sx.gateway.openaiBaseUrl()) using your Silvanexum
API key. Every model call the framework makes is then metered on your own BYO
provider key, attributed, and counted toward your budgets — no duplicate run, no
rewrite. Pass an x-silvanexum-trace header to stitch a multi-step run into one
attributed trace.
The gateway is OpenAI-compatible today (/gateway/openai/v1), so these guides
use an OpenAI model and an OpenAI provider key. Native Anthropic proxying is on
the roadmap.
Guides
LangChain
Route a LangChain chain's model calls through the gateway.
LangGraph
Meter a stateful, branching LangGraph agent.
CrewAI
Meter a multi-agent CrewAI kickoff.
Building with OpenAI Agents SDK, AutoGen, or smolagents? The same pattern applies — any OpenAI-compatible client accepts the gateway base URL. Reach for the Python SDK reference for the full surface.