Notes from running agents in production.

Lessons from 11 million agent runs a month — on autonomy, evals, approvals and the operations work that makes agents trustworthy.

FEATURED

Agent Ops

What we learned running 11 million agent tasks a month

The numbers behind autonomous agents in production: where they fail, how teams earn trust and why the trace matters more than the model.

Ines Calloway

Sep 24, 2026

Agent Ops

Sep 24, 2026

The numbers behind autonomous agents in production: where they fail, how teams earn trust and why the trace matters more than the model.

9 min read

Guides

Sep 10, 2026

A step-by-step playbook for running an agent alongside your team until its decisions are good enough to trust on their own.

7 min read

Engineering

Aug 27, 2026

What we record for every agent run, why each field exists and how traces power replay, evals and audits.

8 min read

Product

Aug 13, 2026

Seat pricing punishes the teams agents help most. Here’s how run-based pricing works and what counts as a run.

5 min read

Agent Ops

Jul 30, 2026

Approvals are the safety net for autonomous agents. Designed badly they become a queue. Here’s how to design them well.

6 min read

Customer Stories

Jul 16, 2026

Lumenpay replaced a deflection chatbot with an agent that issues refunds and closes tickets. Their head of support explains what changed.

6 min read

Field notes

Lessons from running agents in production, new templates and the releases worth knowing about. Unsubscribe in one click.