Every fee, one pour,
onto one slag heap.
SMELT is the coding agent whose fees feed the furnace.97% of every trading fee is poured onto the slag heap.
Read live from Robinhood Chain once the token is launched.
to the buyback, the rest to the builder
0ETH of fees, pours at the threshold
The furnace keeps the job.
Fees land in a crucible that belongs to $SMELT alone. When it fills, the engine pours on its own: 97% buys $SMELT back and throws it on the slag heap, 3% goes to the smith.
It pours at 0.01 ETH. Read live from the engine.
Feed it fees. It keeps burning.
Every trade of $SMELT charges the crucible, and every pour shrinks the supply. The furnace runs for as long as the token trades.
One pour, step by step.
Charge the crucible.
Every buy and sell of $SMELT on Pons drops a little fee into the jar.
0.01 ETH pours a runThe smith's cut.
A small share goes to the wallet that builds and runs the agent.
3%Buy back.
Everything else buys $SMELT, on the Pons curve or on Uniswap after graduation.
97%Onto the slag heap.
What it bought goes to
0x…dEaD. Tokens there can never move again.
Where the fees go
Connecting…Creator 3%Buyback & burn 97%Bounty pool 0%
Appears here at launchFrom the furnace log
each line links to its tx- Nothing yet. The first pour shows up here.
The smith in your terminal.
Tell it what to build or fix. It reads your project, edits files, runs your commands and checks the result. You pay your provider directly. It sells no model access.
Asks before it strikes.
Every edit is shown before it is written, and every command waits for your yes. Add
--yoloto skip the questions.Keys stay on your machine.
Saved in
~/.smelton your computer, or read from an environment variable.Switch models any time.
Name them
provider/modeland change mid-conversation.One command.
The command is
smelt: one standalone file, installed with one line.
Works with the models you already pay for.
- AnthropicCloud
- OpenAICloud
- OpenRouterCloud
- Google GeminiCloud
- GroqCloud
- DeepSeekCloud
- xAICloud
- OllamaLocal, free
- LM StudioLocal, free
- CustomAny OpenAI API
Commands you'll use.
smelt login | Add a provider and choose a default model |
smelt models | List every model your keys can use |
smelt use | Change the default model |
smelt -m | Use another model for one run |
/model | Switch while a session is open |
Agents are just Markdown.
Save this as .smelt/agents/reviewer.md and run smelt -a reviewer.
---description: Code reviewermodel: openrouter/your-modeltools: read_file, search, list_dir---Review the code for bugs and security problems.Do not edit any files.
Light the furnace.
One line installs one standalone file. No Node.js, no account, nothing to sign up for.
$curl -fsSL https://feedsmelt.xyz/install.sh | shcurl -fsSL feedsmelt.xyz/install.sh | shInstall the command line tool. No Node.js needed.smelt loginChoose a provider and paste your key.cd your-projectOpen any folder.smeltSay what you want in plain words.
Get paid to forge SMELT.
Pick a bounty, open a pull request on GitHub, then claim it with your wallet. When the work is approved, the reward is sent to that wallet.
- Loading bounties…
The guild votes.
Vote by signing a free message with your wallet. Each vote weighs as much $SMELT as the wallet holds, all read at the same block when the vote closes.
- Loading proposals…
Heat warnings.
What the burn does and does not do, in plain words.
No promised price.
Burning lowers the supply. It does not set a price; that still depends on how many people want to buy. $SMELT has no intrinsic value and nothing here is financial advice.
Run by an operator.
The engine is a program on a server holding the creator wallet's key, not a smart contract. Every claim, buy and burn is public on-chain, but the operator can change or stop it.
Buys can be front-run.
Each buy, on the bonding curve and on Uniswap v4 after graduation, is limited to a few percent slippage, but bots can still trade around it.
Volume feeds the furnace.
The burn is paid only by trading fees. When trading is quiet, there is little or nothing to pour.