Braintrust (tracing)

Connect your own Braintrust project to receive tracing spans for the executions you run through the Runtype API. This lets you observe your AI executions in your own Braintrust workspace alongside the rest of your LLM tooling.

What you get

Once connected, the executions you run through the Runtype API emit a runtype-execution trace span into your Braintrust project, tagged with execution metadata (user, organization, flow, and execution IDs). Use these spans to confirm which executions ran and correlate them with the rest of your LLM tooling.

This is an execution-level view: the runtype-execution parent span lands in your project, while model-level detail — individual prompt and completion spans, token usage, and latency — is captured in Runtype’s own platform observability. Connecting your project gives you a parallel execution-level copy in your workspace; it doesn’t replace or affect Runtype’s internal observability.

Connect Braintrust

1. Sign up at braintrust.dev and create an API key.
2. In Runtype, go to Settings → Integrations.
3. Find Braintrust and enter your API key.
4. Optionally set a project name — spans land in this Braintrust project. If you leave it blank, Runtype uses a project named Runtype.
5. Click Save.

You can connect one Braintrust integration per organization. Your API key is encrypted at rest.

What gets traced

Tracing fires for every execution you run through the Runtype API — dispatch, product surfaces (chat, API, webhook, and A2A), agent runs, scheduled runs, batches, and evals. Spans are flushed in the background, so a Braintrust outage never slows down or blocks your execution.

Next steps

• What are Logs? — inspect executions inside Runtype
• What are Evals? — test and compare prompts and models
• Connecting AI model providers — manage provider keys