Braintrust
MCPBraintrust evaluates and logs AI application behavior. Antimetal queries experiments and logs to determine whether a prompt or model change caused a regression.
#What Antimetal does with Braintrust
- Query AI experiments and evaluation results
- Read logs and traces of model calls
- Inspect datasets and prompt versions
- Determine whether a prompt or model change caused a regression
#Connecting Braintrust
-
Open Integrations in the Antimetal dashboard
Go to Integrations in the Antimetal dashboard and find Braintrust.
-
Authorize Antimetal in Braintrust
Click Connect and follow the prompt. Depending on the tool this is either an OAuth handoff or a read-only credential you generate in Braintrust and paste into Antimetal.
-
Confirm the connection
Once Braintrust shows as connected, Antimetal begins pulling from it during investigations. No further configuration is required.
#Permissions and access
Antimetal connects over MCP with read-only access and performs no write operations. The exact scopes depend on the credential you issue in the vendor tool — grant read access only.
All tenant data is processed in isolated environments. See security and compliance for details.
Contact us via Slack or at support@antimetal.com.
Related integrations
Feature flags, experiments, rollouts, and product analytics.
Feature flag management for flags, environments, and AI configs.
Product analytics, error tracking, logs, and session replay.
