Agent Runs

Agent Runs lets you inspect eve sessions, turns, tools, reasoning, timing, and token usage in the Vercel dashboard. It appears automatically for eve projects, with no instrumentation file required.

Agent Runs usage is billed at Always-on Tracing rates during beta. All plans include 30-day Agent Runs retention during beta.

Open your project in the Vercel dashboard and go to Agent Runs.

The overview shows:

• Runs over time, broken down by trigger, such as Slack and HTTP.
• Token usage over the same window, split into input, output, and cached tokens.
• A table of runs with the triggering message, trigger type, tokens in and out, turn count, duration, and time.

Select a run to see its model, trigger, and deployment, followed by a per-turn breakdown with:

• Timings for each step in the turn, including skill loads and individual tool calls.
• Input and Output for the turn.
• Reasoning the model produced along the way.
• Tool Calls made during the turn, with their arguments and results.
• Input, cached, and output token counts for the turn.

Agent Runs is eve's built-in OpenTelemetry integration for Vercel. Configure it with agentRuns(). New eve deployments automatically configure tracing to sample at 100%. Existing deployments must configure sampling.

New eve deployments automatically configure tracing to sample at 100%. Existing deployments must add a tracing rule before Vercel collects traces. We recommend sampling at 100% to collect all of the spans related to the run.

For an existing deployment, select your team and project, then open Settings → Tracing:

1. Select Add Sampling Rule.
2. Choose All Environments, Production, or Preview.
3. Set Rate to 100%.
4. Select Save.

Tracing starts when you save the rule. You can also manage sampling rules with vercel traces config. The CLI accepts rates from 1% to 100%. Add a 0% rule from the dashboard.

If your agent processes personal, sensitive, or regulated data, you may need to disclose this capture under applicable laws and in your privacy materials.

The process-wide otel({ tracePolicy }) setting controls the maximum content available to every OpenTelemetry destination. Use recordInputs and recordOutputs to control content capture. Agent Runs cannot restore content excluded by this policy. See Configure process-wide trace content for an example.

Use agent/instrumentation/agent-runs.ts to further narrow what Agent Runs receives. For example, retain spans for private sessions while redacting their inputs and outputs:

import { agentRuns } from 'eve/instrumentation/otel';
 
export default agentRuns({
  exportPolicy: {
    span: ({ audience }) =>
      audience === 'public'
        ? { emit: true }
        : { redact: true, inputs: true, outputs: true },
  },
});

Input redaction removes eve's known prompt, instruction, document, and tool-argument attributes. Output redaction removes response, reasoning, tool-result, exception, and status attributes. Redaction affects only Agent Runs and does not mutate spans sent to another destination.

• Observability: configure process-wide OpenTelemetry settings and third-party destinations.
• Concepts: learn how eve's runtime pieces fit together.
• Vercel Observability: inspect traces alongside logs and metrics.