Replay has always been, and continues to be, pure magic.

Introducing the first post in a new series we’re calling Replay Time Travelogues.

Read @acemarke.dev 's retelling of his time travels through a gnarly React scheduling bug originally shared by @danabra.mov

Get your popcorn ready folks. It’s a real deep-dive!

Replay MCP made this possible because it lets agents use time travel to investigate what happened in the recording. View sources with hit counts, dynamic logpoints that eval every time a line ran, React render analysis with causes. Replay MCP gives agents the same superpowers we give human devs!

Replay Time Travelogue: How Replay MCP Helped Find a React Bug Faster than Dan Abramov Did An example of how Replay MCP enables agents to find and fix deep bugs

Replay time-travel makes solving impossible bugs possible.

Dan Abramov recently fixed a React scheduling bug, but his agent struggled to find it until he added logs to React.

I gave my agent Replay MCP, and it found the bug + fix in <10 minutes!

Details:

blog.replay.io/replay-time-...

but does it come with modules for karaoke or beer? :)

A picture of a carry on suitcase at the airport

Doing the devrel / speaker thing again!

On my way to AI Engineer / React Miami! I'm hanging out all week, and speaking Friday on "React Compiler Rendering".

I'll also be showing off Replay MCP and my recent React tools work, and how they enable agents to fix bugs and improve perf!

Come say hi!

Hey! Flying down tomorrow. See you there!

mark has showed me a preview of this and it's very, very cool

I have some _very_ exciting "bug investigations with and without Replay" results I generated today :)

Will turn these into comparison blog posts next!

(spoiler: Replay recordings enabled fixing seemingly impossible bugs in just a few minutes, exactly as we've been saying :) )

Replay - Browser DevTools from the future. Investigate bugs and flaky tests with perfect reproducibility.

Really excited to see how fast these tools came together, thanks to the React and Redux analysis layers we'd already built as examples. I want to flesh these out with more details, but excited to get the first iterations out there!

They're live in Replay MCP now!

docs.replay.io/basics/repla...

Screenshot of a Markdown format MCP tool describing Zustand stores in a Replay recording

Screenshot of a Markdown format MCP tool describing Zustand state in a Replay recording

Screenshot of a Markdown format MCP tool describing TanStack Query caches in a Replay recording

Screenshot of a Markdown format MCP tool describing a TanStack Query cache entry in a Replay recording

Just shipped the first versions of the new Replay MCP tools for analyzing Zustand stores and TanStack Query caches!

We already had tools for React render insights and Redux actions+state. Now I'm adding more tools to give your agents insights into other libraries as well!

A Replay recording is like nothing else in your software delivery stack. Replay captures what happens at runtime in your web app.

One team said "Now that I see what Replay can do, we're going to add more Playwright tests!"

Go to replay.io to start shipping with confidence.

I’ve posted it before, but it feels evergreen

The two hardest problems in Computer Science are

1. Human communication
2. Getting people in tech to believe that human communication is important

read "react" and "backbone" together in the same sentence and got very confused for a minute :)

Replay recordings make infinite runtime details available. Humans can investigate. Agents can do it faster and with more patience.

I plan on putting up a blog post about this in the next couple days with the details on this session!

In this case, my own agent:

- Looked at React renders
- Identified paired Suspense renders each keystroke
- Last update didn't happen in prod
- Added logpoints to React internals
- Traced line hits and fiber values
- Confirmed last update bailout and what logic needed to fix it

Both mattered.

Dan already had the repro. But, he had to have his agent add logs to React, rebuild, redeploy, read logs, repeat.

Replay has "logpoints" - dynamic logging, eval'd every time a line of code ran. No need to rebuild! Record once, dig as deep as needed.

Replay recordings make infinite runtime details available. Humans can investigate. Agents can do it faster and with more patience.

I plan on putting up a blog post about this in the next couple days with the details on this session!

In this case, my own agent:

- Looked at React renders
- Identified paired Suspense renders each keystroke
- Last update didn't happen in prod
- Added logpoints to React internals
- Traced line hits and fiber values
- Confirmed last update bailout and what logic needed to fix it

Both mattered.

Dan already had the repro. But, he had to have his agent add logs to React, rebuild, redeploy, read logs, repeat.

Replay has "logpoints" - dynamic logging, eval'd every time a line of code ran. No need to rebuild! Record once, dig as deep as needed.

Replay - Browser DevTools from the future. Investigate bugs and flaky tests with perfect reproducibility.

Here's the Replay MCP docs:

docs.replay.io/basics/repla...

You might have seen me excitedly posting about my recent improvements to Replay MCP to add better React analysis:

bsky.app/profile/acem...

I'd love to have folks try this out, and also answer questions about it! What is it, how does it work, benefits, etc.

Ask me anything about Replay MCP!

Both!

We've seen agents fix problems they wouldn't have solved otherwise, _and_ fix problems faster.

Great example here:

bsky.app/profile/acem...

Dan's agent "burned a ton of tokens" over a month.

I fed recordings to my agent yesterday. Cause + fix in <10min, <$10.

Replay - Browser DevTools from the future. Investigate bugs and flaky tests with perfect reproducibility.

Try it out - would love to chat about how Replay MCP can help your agents fix bugs, fix flaky tests, and improve app perf!

docs.replay.io/basics/repla...

Replay MCP really does make a big difference in how agents can debug code... and we're just getting started!

I've personally built a bunch of the new MCP tools for analyzing React renders, Redux actions and state, and more, and it's already drastically improved the results!

bsky.app/profile/acem...

docs.replay.io/basics/repla...

Building Replay MCP - giving time travel debugging superpowers to agents!

Some examples:

bsky.app/profile/acem...

Heyyyy @danabra.mov ! I just created Replay recordings of your useDeferredValue bug . Fed them into a fresh agent session with general "issue report, go investigate" instructions...

and Claude + Replay nailed the cause + solution in <10 minutes!

Will write this up later, but this is great!

Heyyyy @danabra.mov ! I just created Replay recordings of your useDeferredValue bug . Fed them into a fresh agent session with general "issue report, go investigate" instructions...

and Claude + Replay nailed the cause + solution in <10 minutes!

Will write this up later, but this is great!