The flight recorder for coding agents.
Postrun records every command, edit and file your agents touch, on your own machine. Review the whole session, then share a redacted report when someone else needs to see it.
Local by default. Nothing leaves your machine unless you export it. See an exported report
An example Claude Code session streaming into postrun, step by step: a failed test that the agent then fixed, and AWS credentials the agent printed, masked as [REDACTED].
An example Claude Code session streaming in. The agent printed AWS credentials; Postrun masks them before anything is shared.
Your agent finished.
What did it actually do?
The chat scrolls away. The diff shows the end state, not the three approaches it tried, the test it broke, or the credentials it printed on the way. Postrun keeps the full record: every step, in order, with its output.
Capture on your machine
Hooks and local telemetry feed a recorder bound to 127.0.0.1. Prompts, tool calls, output and edits land in a private store in your home folder.
$ Read the whole session
One timeline across agents: every turn and step, the files it touched, the commands it ran, what failed and why. It updates live while the agent works.
$ Send a report, not your logs
Export one session as a single HTML file. Secrets, credentials and home paths are masked first, and you see every masked value before you send it.
$ The session is the unit. Everything else is a filter.
Live as it runs
Watch steps land while the agent works. Close the tab, come back, and the session has caught up.
Every agent, one schema
Adapters turn each agent's own logs into the same steps, turns and sessions, so you review them the same way.
Edits as diffs
See each edit the agent made, including the ones it later undid, not just the final state of the branch.
Failures up front
Failed commands, error types and exit codes are flagged in the timeline and counted in the summary.
Bring your own agent
A small authenticated ingest API takes v1.2 events from any adapter, validated before anything is stored.
Redaction you can check
Keys, tokens, passwords and home paths are masked on export, and each one is listed so you can review it.
Your sessions hold your secrets. They stay with you.
Agent sessions contain full prompts, tool output and anything a command printed. Postrun is built so none of it leaves your machine by accident.
127.0.0.1The recorder and the review UI bind to loopback only and refuse other Host headers.chmod 600Captures and the session store are owner-only on disk.no telemetryPostrun never phones home. Nothing syncs in the background.export onlyA session leaves your machine only when you export it, one at a time.[REDACTED]Secrets and credentials are masked on export, and you review every mask.CSP: noneReports run no scripts and load nothing, so a hostile agent output stays inert.Before you ask.
How Postrun handles your sessions, your secrets and your team.
01What does Postrun record?
Everything a coding agent does in a session: your prompts, its replies, every command it runs with the output, every file it reads or edits, and which steps failed and why. Each session becomes one timeline of turns and steps you can review from start to finish.
02Does any of my code, prompts or output leave my machine?
No. The recorder and the review app run on your computer and listen on 127.0.0.1 only. Sessions are stored in ~/.postrun, readable only by your user account. Postrun has no telemetry and never phones home.
A session leaves your machine only when you export it yourself, one session at a time.
03Which coding agents does it work with?
Claude Code and Cline today, recorded into one shared format so you review them the same way. Cursor and Codex are next. Any other agent can send sessions through Postrun's local ingest API.
04How does sharing work?
Export a session and you get a single HTML file: every command, edit and message, expandable, with failures and output. It has no JavaScript and loads nothing from the internet, so it's safe to open and works anywhere: email it, share it alongside a pull request, or drop it in a chat.
Hosted share links are planned, built on the same redacted file.
05What does redaction catch, and can I rely on it?
Before anything is exported, Postrun masks known secret formats (private keys, AWS, GitHub, Anthropic, OpenAI, Stripe, Slack and Google keys, JWTs), passwords in URLs, Authorization headers, and values assigned to names like API_KEY, token or password. Home folder paths become ~, and your machine name is left out.
Every masked value is listed for you to check before sending. Redaction is a safety net, not a guarantee, so skim the report before you share it.
06Is Postrun a cost or usage dashboard?
No. It shows the cost and token counts your agent reports, but the point is review: what the agent did, what it touched, and what went wrong. The session is the unit, and everything else is a filter.
07Can I watch a session while the agent is still working?
Yes. The review app updates live as steps are recorded, and catches up on its own if you close the tab and come back.
08Where is my data, and how do I delete it?
Everything Postrun stores lives in ~/.postrun on your computer. Delete that folder and it's gone. There's no account and no copy anywhere else.
09How do I get it?
Postrun is in early access. Join the waitlist and we'll email you the install when your agent is supported.
Know what your agents did.
Postrun is in early access. Leave your email and we will send you the install when your agent is supported.
Works with Claude Code and Cline. Tell us which agent you use.