the how-it-works for AI-written code

the agent ships features.
you keep the thread.

Your code changes every day. Your understanding shouldn't fall behind. Backthread keeps a running log of what the AI decided and why, on a live map of your system.

works inside Claude Code · Cursor · Codex · Gemini

what you get

Three ways to keep the thread.

AI writes most of the code. You can outsource the thinking, but not the understanding.

01 · just ask

Ask it, in plain English

Ask "how does login work?" in your agent. You get a short answer, with links to the decisions behind it. No grep, no re-reading files.

02 · the why

Every decision, written down

Your agent makes dozens of decisions a day. Backthread records each one and why, and flags the ones you never approved.

03 · the what

The whole system, on one map

A live map, built from your code. Click a decision to see where it landed. Hover any box for its history. Drag the slider back in time.

Ask how it works.

A high-level view of every decision, in the browser or in your agent via MCP. You get deterministic answers about how things work.

claude code · your-app
> /backthread:how why do we sync to hubspot at night?
The daytime API kept hitting rate limits, so your agent moved the customer → HubSpot sync to a nightly batch at 02:00 UTC. Trade-off: the CRM can be up to a day old. The team decided that's fine.
3 decisions cited · customer sync → hubspot · captured from Codex, 2 weeks ago
MCP · why asking beats reading

The "why" isn't in your code. Stop making your agent read all of it.

100×+ fewer tokens per "why" question. The MCP answers on our server, so your agent doesn't grep the whole repo.
1 call instead of dozens of file reads that still wouldn't find the reason.
0 lines of your code leave your machine. Only the plain-English "why" is sent.

A "how does X work?" used to cost your agent thousands of tokens of reading. The answer isn't even in the code. Backthread returns it in one call.

for teams

Everyone pulls their own thread. Backthread holds the whole one.

On a team, no one sees every decision. Your teammates' agents ship code you never reviewed, and the reason for a change is never in the diff. Backthread shows the why to everyone.

// transparency

See why any agent did anything

Every decision, from anyone's agent, on one shared map with the why attached. Onboarding, on-call, or reviewing a PR: the reason is right there, not stuck in someone else's session.

// above the diff

Read the decisions, not every line

A red dot marks the decisions no one approved. Reviewers focus on the few that change the architecture, instead of rubber-stamping PRs no one fully read.

// stay in the loop coming soon

A weekly email digest

An email of what your team's agents decided each week, so no one has to rebuild the picture from PRs.

you're not imagining it
"Comprehension debt is the growing gap between how much code exists in your system and how much of it any human being genuinely understands."
Addy Osmani · Google · O'Reilly, 2026
"If an LLM wrote every line of your code, but you've reviewed, tested and understood it all, that's not vibe coding, that's software development."
Simon Willison · creator of Django
"You fully give in to the vibes, embrace exponentials, and forget that the code even exists."
Andrej Karpathy · ex-OpenAI, coined "vibe coding"
the payoff

What you stop doing.

Stop digging through old code.

Every bug used to start the same way: what is this file, why is it like this? Now it starts with an answer.

Stop reading every PR to stay in the loop.

Backthread flags the decisions that matter. Read the few that count, skip the rest.

Stop relearning your own system.

It was recorded the whole time. You just had nowhere to see it.

it's real

There's a live demo. Go try it.

Not screenshots. It's live. Click a decision to light up the code it touches, read the why, drag the slider through history. Built from a real repo.

Open the demo →
get started

One command. Any agent.

See it on your own repo, free. Pick where you code:

/plugin marketplace add backthread/backthread
/plugin install backthread@backthread

Then run /backthread:start. Same setup, built into Claude Code.

npx backthread install --agent codex
npx backthread install --agent cursor
npx backthread install --agent gemini
not in a terminal? start in your browser →
pricing

Free to see it. Cheap to keep it.

Build your repo once, free. After that, a flat price per seat with a monthly allowance included. Go over it and you pay a clear per-unit rate.

 

Free

A real build of your own repo. Once.

$0
  • One full build of one repo
  • Your real map + decision log
  • Ask it anything, once it's built
for founders

Solo

For when you are the whole eng team.

$19 /mo
  • 1 seat · 1 repo
  • Rebuilds itself on every merge to main
  • Full history + time slider
  • Ask it anything, any time
for teams

Team

For the 2–30 people losing the thread together.

$25 /seat/mo
  • Seats you add and drop, GitHub-style
  • Everyone's decisions on one shared map
  • Shared answers across the whole team
  • 300 rebuilds + 500 sessions per seat

Extra repos: +$10/mo each; you get one to start. Over your monthly limit? $0.20 per rebuild, $5 per 1,000 sessions. Most people never get close.

security

Your code stays yours.

Backthread connects to two things: your GitHub repo, to build the map, and your coding agent, to capture the why. Here's what each one does with your code.

// github integration

Reads your code to build the map

The GitHub App copies your repo into a temporary sandbox, reads its structure, and deletes it. Your code is never saved to our database. Access is read-only, one repo at a time, and you can turn it off in GitHub anytime. Private repos stay private.

// agent integration

Reads your sessions to capture the why

The plugin reads your agent's chat on your own machine, removes the code and command output, and sends only the plain-English "why." Your code itself is never sent. It's open source, so you can check it →

The full list of what leaves your machine, what we store, and every sub-processor is on /security.

questions

Things you might be wondering.

Can I really just ask it questions?

Yes. Ask "how does X work?" or "why is it built like this?" in your agent. You get a short answer in plain language, with links to the decisions behind it. It uses the full history, so the answer is current.

I'm shipping fast right now. Do I need this?

Probably sooner than you think. The pain isn't loud on day one. It's slowly losing the ability to say what your system does without re-reading it. Most people hit it around week three of heavy AI use: every bug needs a research session first. Backthread keeps the picture current, so you don't rebuild it by hand.

How is this different from a code-review tool?

Code review shows what changed in one PR. Backthread shows what your whole system is now, what it was a month ago, and why it changed. Different question, above the diff.

Do you read or store my code?

We copy your repo into an isolated sandbox, read it, save only the map and decisions, then delete the sandbox. Your code never lands in our database, logs, or backups. The plugin also reads your agent's transcripts, but removes the code and command output on your machine first. Full details on /security.

Who's building this?
Jevgeni Bogatyrjov
Jevgeni Bogatyrjov linkedin.com/in/bogatyrjov →

Solo founder so far, building in the open. I use the same agents Backthread is built for, which is how I found the problem. It's live and free to start at app.backthread.dev. I'm looking for the first people to run it on a real codebase and tell me what's missing.

ready?

Stop re-reading
your own code.

Point Backthread at your repo. Ask it anything, see every decision, read the why behind each one.