Using asciinema to record coding-agent sessions for replay debugging

On 2026-11-07 at 17:01, the AI section of DEV Community published a piece titled Recording coding agent sessions with asciinema and replaying them for troubleshooting. The article walks through using asciinema 3.x to capture the terminal output of coding agents such as Claude Code or Codex, timestamps included, into replayable and searchable text files, so that after the fact you can pinpoint which step an agent got wrong.

The scene it sets up is familiar: the agent claims the flaky test is fixed, and the suite agrees. Somewhere in the few hundred lines it printed while you were on another tab, it wrote one sentence about why. You scroll back up to find it, but the interface has redrawn itself many times since, and the line you want is either gone or broken off. From memory you end up guessing, or re-running things in the hope of catching it.

The piece also warns that a recording contains every byte the agent prints, including potential secrets, so files should stay local. It appears as part of the AI Agents Pocket Guide, a series of 8 short books covering agents, RAG, prompts, observability, events, databases and system design. The same page lists the author's project Hermes IDE, aimed at developers who ship with Claude Code and other AI coding tools, plus links to xgabriel.com and GitHub.