Review an agent's work before you merge.
A pull request shows where the branch ended up. Postrun shows how it got there, so you can review the work and not only the result.
The problem
When a coding agent opens a pull request, the diff is the end state. It doesn’t show the commands the agent ran, the test that failed twice before it passed, the files it read and changed back, or the credentials a command printed on the way. The chat that would tell you scrolls away, and the agent’s own summary is the agent describing itself.
So the reviewer either trusts the summary or reconstructs the session by hand. Neither is a review.
How Postrun handles it
- STEP 01
Record while the agent works
Postrun captures Claude Code and Cline sessions on your own machine as they run. Prompts, replies, tool calls, command output and edits land in a private store in
~/.postrun. You keep working the way you already do. - STEP 02
Open the session, not just the diff
The review app lays the session out as one timeline of turns and steps. Above it: a summary of steps, files touched and commands run, a list of every file the agent created, edited or read, and a table of the commands it ran with their exit codes. If the agent is still working, the view updates live.
- STEP 03
Start with what failed
Failed commands are flagged in the timeline with their error type and exit code, and counted in the summary. Expand one to read its output, then follow the next steps to see whether the agent fixed the problem or worked around it.
- STEP 04
Read the edits as diffs
Every edit the agent made is shown as a diff in the order it happened, including edits it later undid. You see the approaches it tried, not only the one that survived into the branch.
- STEP 05
Send the report with the pull request
When someone else is reviewing, export the session. Postrun masks secrets, credentials and home paths, lists every value it masked so you can check them, and writes one HTML file that opens in any browser with no account and no install.
What a reviewer can check
- Commands you didn’t expect the agent to run, with their full output.
- Tests or builds that failed, and whether they were run again afterwards.
- Files touched outside the scope of the change.
- Edits that were made and then reverted, which the final diff hides.
- Secrets that appeared in output, which are masked and listed when you export.
All of this stays on the machine that recorded it. Nothing leaves unless you export a session, one at a time. See security for the details.
Review the session, then merge.
Postrun is in early access for Claude Code and Cline. Open the example report to see exactly what an export looks like, or join the waitlist for the install.