Eve

​

Setup

pnpm add braintrust eve

npm install braintrust eve

BRAINTRUST_API_KEY=<your-braintrust-api-key>

​

Manual instrumentation

• Eve 0.62 or later: Add one instrumentation file. See Eve 0.62 or later.
• Eve versions before 0.62: Add a hook and a root instrumentation.ts file. See Eve versions before 0.62.

​

Eve 0.62 or later

import { braintrustEveInstrumentation, initLogger } from "braintrust";

export default braintrustEveInstrumentation({
  metadata: {
    app: "my-eve-agent", // Replace with your app name
  },
  setup: ({ agentName }) => {
    initLogger({
      projectName: agentName, // Replace with your project name
      apiKey: process.env.BRAINTRUST_API_KEY,
    });
  },
});

eve build && eve start

​

Eve versions before 0.62

import { braintrustEveHook } from "braintrust";
import { defineState } from "eve/context";
import { defineHook } from "eve/hooks";

export default defineHook(
  braintrustEveHook({
    defineState,
    metadata: {
      app: "my-eve-agent", // Replace with your app name
    },
  }),
);

import { braintrustEveInstrumentation, initLogger } from "braintrust";
import { defineState } from "eve/context";
import { defineInstrumentation } from "eve/instrumentation";

export default defineInstrumentation(
  braintrustEveInstrumentation({
    defineState,
    setup: ({ agentName }) => {
      initLogger({
        projectName: agentName, // Replace with your project name
        apiKey: process.env.BRAINTRUST_API_KEY,
      });
    },
  }),
);

eve build && eve start

​

What Braintrust traces

• Turn spans (eve.turn), as individual traces, with the user message as input, the agent’s final response as output, and the session ID in metadata as eve.session_id. With the setup for Eve versions before 0.62, turn spans also have aggregated token metrics.
• Parent-child nesting under the caller’s span, with the setup for Eve 0.62 or later, when the caller propagates W3C trace context to Eve. Eve forwards the caller’s traceparent header for session requests that include a remote subagent callback. Requires braintrust v3.37.0 or later.
• Step spans (eve.step), as LLM spans nested under their turn, with the full message history as input, the model response as output, and model name and provider in metadata.
• Tool call spans (named after the tool), as tool spans nested under their turn, with the tool arguments as input and the tool result as output.
• Subagent spans (named after the subagent), as tool spans nested under their turn, with the subagent’s turn spans nested under them.
• Token usage metrics (prompt, completion, total, prompt cached, and prompt cache creation) on step spans. With the setup for Eve versions before 0.62, step spans also have estimated cost when Eve reports it.
• Errors captured on every span.

​

Resources

• Eve documentation
• eve on npm
• Braintrust tracing guide