Integrations
Integrations
Every way to send answers to Latent and every place Latent sends alerts, with the plans that include each one.
Latent connects at two ends: where your answers come in, and where its alerts go out. Every integration is documented here whatever your plan, and each one is marked with the plans that include it.
Ways in
| Integration | What it does | Status |
|---|---|---|
| Python SDK | Captures every finished call your app makes through the OpenAI, Anthropic or Google GenAI Python SDK and sends the prompt and answer for scoring. One line at startup. | Every plan |
| Coding agent setup | One prompt that has Claude Code, Codex, Cursor or another coding agent add the SDK to your project and check the first answer arrives. | Every plan |
| Proxy | An endpoint your app sends its model calls through, so answers are scored with no SDK in your code. | Roadmap |
| Observability sink | Scores the answers in traces you already export to Braintrust, Langfuse or an OpenTelemetry collector. | Roadmap |
When Latent runs in your environment, it reads your model inside vLLM, or re-reads each finished answer with the sidecar. Deployment options covers both.
Ways out
| Integration | What it does | Plans |
|---|---|---|
| Slack | Posts each alert to a channel through an incoming webhook | Pro and above |
| PagerDuty | Opens an incident when a rule fires and resolves it when the rule clears | Pro and above |
| Sends each alert to up to 20 addresses | Pro and above | |
| Webhook | POSTs the full alert as JSON to an endpoint you run | Pro and above |
Alerts and integrations explains the rules that decide when an alert fires, how to set up each destination, and what each one receives.
Was this page helpful?
Updated 3 October 2026