Record what you build. Export what you learned.

Why I built it

I wanted a simple way to record commands so I could showcase other CLI applications I was building. I somehow did not know about asciinema at the time, which is both funny and very on-brand, but the original itch was real: something straightforward, local, and easy to export into whatever format I needed.

Then the idea grew. If humans need better terminal recordings, AI agents definitely do too. When an agent builds a CLI app, it runs commands, things change, things fail, output scrolls past. Reviewing that work usually means trusting memory, scattered logs, or “trust me bro, the agent ran it.” I wanted the whole session to have a durable shape.

What it does for AI agents

This is the part I care about most. Rekord captures what an agent ran, what changed, what failed, and what actually appeared in the terminal, then hands that back as structured, AI-ready context through MCP tools.

So another agent (or a human) can review the run, reproduce the flow, and understand what happened without reconstructing it from vibes. It gives terminal work a memory.

What it does for humans

Same engine, different output. A session becomes:

• a replayable cast
• docs of the commands and output
• a video for demos and showcases

I use it to record command walkthroughs, show off terminal apps I’m building, and replay what happened in a past session.

Install

# Homebrew
brew tap Omotolani98/rekord
brew install rekord

# or from source
go install github.com/Omotolani98/rekord/cmd/rekord@latest

Verify:

rekord version

Record a session

rekord start --name my-demo
# run your commands normally, stop with Ctrl-]

One-shot, no session:

rekord run --name demo -- make build

How it works

Rekord wraps a terminal session, records commands and output, and keeps enough structure around the run to make it useful after the fact. Locally, with zero telemetry, safe by default. MIT licensed.

The goal is not to replace your terminal. It’s to give terminal work a memory. Run the thing, record the thing, export the thing, and let humans or agents understand what happened.

Source on GitHub →