LangGraph
Discover the agents deployed on a LangGraph server.
Airia discovers the assistants (agents) and their graphs on a deployed LangGraph server and adds them to the inventory.
Authentication model
A LangSmith API key plus the URL of your LangGraph deployment.
Two things to get right, because they are the most common setup mistake:
- The URL is the deployment's own server URL, not the LangSmith website. Airia calls the
LangGraph Server API directly (it searches
/assistantson that host). A LangGraph Platform deployment URL looks likehttps://<deployment-name>.<region>.langgraph.app; a self-hosted server is whatever host you run it on.https://api.smith.langchain.com(the LangSmith control plane) is not a valid value — it returns 404 on the assistants API and the connection test fails. - The API key is a LangSmith API key (starts with
lsv2_pt_for a personal token orlsv2_sk_for a service key), created under LangSmith Settings → API Keys. Airia sends it to the deployment as thex-api-keyheader.
Customer-side setup steps
- Identify the LangGraph deployment you want discovered and copy its server URL (from the LangSmith Deployments page, or your self-hosted host).
- In LangSmith, Settings → API Keys → Create API Key (Personal Access Token is fine),
copy the
lsv2_...value. - In Airia, go to Discover → Connections → LangGraph and create a new connection with the deployment URL and the API key.
Form fields the Airia UI asks for
| Field | Required | Notes |
|---|---|---|
| URL | Yes | Your LangGraph deployment server URL (not the LangSmith website) |
| API key | Yes | A LangSmith lsv2_... key. Stored encrypted, masked on edit |
Multiple environments
Multiple LangGraph connections per tenant are supported: one per deployment.
Was this page helpful?