Point it at a token on Robinhood Chain. Four lanes run in parallel: what the contract can still do to you, how deep the pool is, who deployed it, and how few wallets hold it. Every check is a contract call, so a run costs nothing and needs no key. It never predicts a price.
A loop over sample launches, so you can watch the shape work. graphrun reads real chains on your machine; this page reads nothing.
Python 3.10 or newer is the only thing you need in advance. Steps one to four need no key and no account. Step five adds the two lanes that want an index.
Two dependencies, PyYAML and certifi. Nothing is sent anywhere.
git clone https://github.com/hanakoxx/graphrun.git cd graphrun pip install -e .
Draws the graph in your terminal and names the return path you did not build. Run it on a chain config and it will tell you the graph is fast and never gets smarter.
graphrun explain examples/pons-live.yamlThe real graph against a deterministic backend. One lane fails on purpose so you can watch a null land in the merge instead of killing the run.
graphrun run examples/pons.yaml --dry
Put the token address in the config, or build one below. Two lanes read the chain directly and work right now, with no key at all.
graphrun run examples/pons-live.yamlWho deployed a contract and who holds it need an index. Free tier at dev.blockscout.com, one minute, no card.
export GRAPHRUN_EXPLORER_KEY=proapi_...
graphrun run examples/pons-live.yamlDefaults point at Robinhood Chain. Override either endpoint with an environment variable and the same graph reads any EVM chain.
Inside the contract lane, these are the four bytes it looks for in the deployed bytecode. Every one is computed with keccak256 at runtime, never copied from a table, because a wrong constant fails silently: the call reverts, the lane reports nothing, and nothing reports nothing.
Give four checks one shared window and they converge: the first finds something, the rest read it, and all four reports centre on the same thing. Each lane here sees only its own slice, which is what makes them return different things.
Can the owner still change the rules after you are in. Scans the deployed bytecode for mint, fee switches, blacklists and pause, then checks whether ownership was ever renounced.
Pool reserves, total supply and decimals, read straight from the contracts. The lock check reports not verified rather than guessing, because reading a lock needs the locker contract and that is not wired yet.
Who shipped it, how many transactions that wallet has sent, and whether it holds anything. Half a million transactions with an empty balance is a factory, not a person, and that changes what the answer means.
How concentrated the supply is, and whether a single wallet holds more than half of what the explorer lists. Concentration is arithmetic, so it costs nothing once the list is in hand.
On a launch, how bad is this mistake has exactly one translation: can you sell. Move the rules and watch the same findings change sides. This is the real gate logic, running here.
A sample launch, so the rules have something to bite on. Turn a lane off to see what a graph does when one node times out.
Neither of those is a threshold set high. They are checks that never pass, no matter how the chart looks. Confidence is the weakest input to this decision, because it is the only one anyone can influence.
This runs in your browser and touches no network. Drop in a token address and it writes the graph for you, ready to save as examples/my-launch.yaml and run.
Nothing can, and a tool that implies otherwise is worse than no tool at all. This reads what is in front of you and whether you could get back out. Everything after that is your call. It holds no keys, signs nothing, and never places a trade.
The tool is MIT and works with or without the token. Everything below is read from the chain the same way graphrun reads anything else, and you can run the same four checks on this contract yourself in one command.
Verify the contract address on the explorer before you touch it, and run the four checks on it the same way you would on anything else. That is the whole point of the thing.
published at launch