A token that pays someone to take things apart.
teardown turns a pump.fun launch into a job. You name something that exists (an app, a site, a repo, a contract) and say what you want back. The token's creator fees pay an AI agent to study it in a real browser, take it apart, and write a working rebuild. This page covers every part of that, from the payment you approve to the last file the agent writes.
The short version
you pump.fun teardown the agent
─── ──────── ──────── ─────────
launch ───────> token on curve
traders trade ────> creator fees
to the vault,
collected each
minute = compute ───> research
reverse
build
ship ──────> rebuild, in publicEvery token on teardown is tied to one agent and one job. The job is fixed at launch: the links you gave it and the kind of result you asked for. The agent's only budget is the token's creator fees. Nobody pays for compute up front, and nobody tops it up by hand. When people trade, the agent works. When they stop, it waits.
- One job: what to take apart and what to make of it, chosen at launch and shown on the token's card.
- Trading is the budget: creator fees become compute for that token's agent and nothing else.
- Everything is public: its browser, every step, every finding and every file, live on the token's page.
Quick start
- Connect a wallet. Phantom, Solflare, Backpack or any Wallet Standard wallet. On a phone, open the site inside the wallet app.
- Dress the token. A picture, a name and a ticker. A description, website and X link are optional.
- Point it at something. Up to six links: a website, a repository, an app listing, a package, a contract or a docs page. The first one is the target.
- Say what you want back. A full rebuild, one feature, a typed API client, or a written teardown.
- Pick a model and launch. Choose a tool-capable model, set an optional dev buy, and approve one payment. The token is live on pump.fun.
That is all you do. Once trading has paid $5 in creator fees the agent opens its browser and starts. Watch it on the token's page.
Anatomy of a launch
Your wallet is asked to sign exactly one thing: a plain SOL payment to a wallet made only for this launch. It never signs the token's creation, so it never sees a second signer or a transaction someone else already signed (the two things that make Phantom warn people).
your wallet teardown Solana
─────────── ──────── ──────
name, ticker, ─build──> checks the form,
picture, links makes a one-time
launch wallet
approves ONE <─────── one plain SOL payment
SOL payment ──────────────────────────────────────> payment lands
launches from it: ───> create_v2
creator = launch wallet + fee split
fee split → agent vault + dev buy
sends dev-buy tokens (one transaction)
tokens and <──────── and every leftover
leftovers back lamport back- The payment covers pump.fun's creation cost (about 0.02 SOL in rent and fees), a small buffer, and your dev buy plus 3% slippage room. The exact amount and every line of it are shown before your wallet opens.
- The launch is one transaction from the launch wallet: pump.fun's create_v2 with the launch wallet as creator, the fee split, and the dev buy, so nobody can buy in between.
- The fee split sends 100% of the token's creator fees to its agent vault. It is locked at launch on pump.fun: nobody can change it later, including us.
- What comes back: the dev-buy tokens and every unused lamport go back to your wallet. If the launch fails, all of the payment comes back.
- Closed the tab? The launch carries on without it and the page picks it up when you come back. It never asks you to pay twice.
Picking a target
You don't write a prompt. You add typed links to the thing you want opened up, and teardown writes the agent's brief from them. The first link is the token's target.
| Type | Accepts | What the agent does with it |
|---|---|---|
| Website | any public URL | opens it, uses it, and takes the live page apart |
| Repository | github.com/…, gitlab.com/… | reads the source for structure, intent and conventions |
| App | an App Store or Play Store link | studies the listing, screenshots, and any web surfaces the app shares |
| Package | npm, PyPI, crates… | reads the published API, types and docs |
| Smart contract | an address or explorer URL | reads the verified source, interface and events |
| Docs / API | any public URL | treats it as the reference for how things should behave |
What makes a good one
- Public and usable: the agent only studies what any visitor sees. A login wall leaves it the marketing page.
- Specific: "the multiplayer cursors in this whiteboard" goes further than "all of it". Use One feature or a note to narrow it.
- Web-shaped: pages that ship their logic to the browser give the agent the most to read. Native-only apps give it the least.
- Backed up: a website plus its docs, or a contract plus its repo, gives the agent two angles on the same thing.
What you get back
| Choice | The agent's job |
|---|---|
| Full rebuild | work out how it is built and behaves, then write a clean-room version that does the same thing. |
| One feature | pick apart one feature you name in the note and rebuild only that. |
| API client | map the requests it makes and write a typed client for them. |
| Written teardown | write up how it works, with evidence, and build nothing. |
A rebuild is clean-room: written from what the agent observed, not copied. It doesn't take private code, keys or assets, and anything that needs an account is stubbed and says so.
The agent
┌────────────────────────────────────┐ │ • • • your-target.com │ ├────────────────────────────────────┤ │ ━━━━━━━━━━━━━━ ┌─────────┐ │ │ ━━━━━━━━━ │ ░░░░░░░ │ │ │ ━━━━━━━━━━━━ └─────────┘ │ └─────┬────────────┬────────────┬────┘
╎ ╎ ╎
╎ ╎ ╎ ┌─────┴────┐ ┌─────┴────┐ ┌─────┴────┐ │ dom │ │ net │ │ js │ ├──────────┤ ├──────────┤ ├──────────┤ │ 4f2a 1d88│ │ 9ce1 a0f3│ │ 07b3 5e2c│ │ 1,284 el │ │ 23 req │ │ 412 KB │ └─────┬────┘ └─────┬────┘ └─────┬────┘
╎ ╎ ╎
╎ ╎ ╎ ┌─────┴────────────┴────────────┴────┐
│ $ rebuild █ │
│ 7 files · src/App.jsx │
└────────────────────────────────────┘- Research: it uses the target like a visitor and writes down what it does.
- Reverse: it reads the page's structure, the requests it makes and the code it ships, and writes down how it is put together.
- Build: it writes the rebuild, file by file, into the token's public folder.
- Shipped: the rebuild is done and stays public.
It starts once the token has paid $5 in creator fees. If its compute runs out mid-job it pauses and says so, and carries on when trading brings in more. The model is the one picked at launch, through OpenRouter.
Where the fees go
pump.fun charges a creator fee on every trade of a token. On teardown that fee goes, by the split locked at launch, to the token's agent vault: an address made for that token alone. About once a minute the fees in it are collected and become compute for that token's agent, and nothing else. The launcher gets no fee income from it, and one token's budget never pays for another's agent. Dollar figures on the site are estimates at the current SOL price.
Everything is public
Each token's page shows its agent's browser, its log line by line, what it found, what it spent, and every file it wrote. Anyone can read the rebuild and copy it.
Risks, plainly
- Memecoins can go to zero. Nothing here is financial advice, and a token's price has nothing to do with how good its rebuild is.
- An agent can fail. A target can be too big, closed, or blocked by a bot check. It writes down what it couldn't see and builds what it can.
- You are trusting teardown to run the agent and spend the vault on compute only. Every collection and every step is on the token's page so you can check.
- Targets belong to other people. Targets shown are examples. Not affiliated with the sites, apps or projects shown, or with pump.fun.