Why we’re building Glint
We know everyone’s posting a thesis right now, so here’s ours. We think trading agents will create a new market for something traders have always valued but had a hard time buying: someone else’s way of trading. A trader who’s spent years refining a strategy should be able to turn it into an agent, show people how it trades, and earn when someone wants to run a copy. That’s why we’re building Glint.
Think about how trading ideas travel today. Someone posts that they’re buying ETH on the way down. You see the entry, maybe an update a few days later. What’s harder to pick up is how they sized that first buy, why they waited before buying again, or why they stopped while they still had cash. You can agree with the original call and end up trading it completely differently. A lot of a trader’s skill lives in those decisions.
What caught our attention when @Lighter_xyz brought its AI Agent Kit to Robinhood Chain was the ability to test a strategy against a live order book. You could watch an agent make those decisions as the market moved, paper trade it, and decide when you were comfortable putting money behind it. That got us thinking about both sides of the trade. What would make a trader confident enough to publish an agent? And what would another trader need to see before buying it?
On Glint, you start by describing your strategy in your own words. You choose which assets your agent can trade, how much it can spend, and how much it can hold. It can trade ETH, tokenized stocks, and ETFs on Lighter’s spot markets on Robinhood Chain. The agent proposes an order, and Glint checks it against your limits before sending it. If you’ve capped your ETH position, the agent has to respect that cap even when its next decision is to buy more.
The marketplace is the part we’re most bullish on. Before you buy an agent, you can read its full strategy and look through its onchain trading history. Say you’re comparing two ETH agents after a selloff. Both finished the week around the same place. One spent its budget on the first drop and sat through the rest. The other bought in smaller amounts over several days. Seeing that difference tells you something a return figure alone never could. You can decide which approach you’d actually want to run.
For the creator, that same history becomes part of what they’re selling. They can list the agent they’ve been running, and buyers can judge the work for themselves. Each time someone buys a copy, the creator earns from the sale.
Long term, our bet is that people will follow agent builders the way they follow traders today. If you’ve run someone’s ETH agent for six months, you have a pretty good reason to pay attention when they publish their next strategy. We think entire businesses will grow out of that relationship. We’re building Glint to be the place on Lighter where a trader can find their first customer and give them a reason to come back.
https://t.co/ZRdPlNsgaX
https://t.co/Fothp4v3NV
Lighter AI Agent Kit now supports Robinhood Chain!
Test your strategies against live order books, monitor their performance with the paper trading engine, and deploy them to production.
Compatible with your favorite harness.
Install in one line!
https://t.co/XM5A7Y4IuC
If the model times out or returns something invalid, it gets one retry with the error included.
If there's still no valid answer, the agent doesn't guess. It holds that check, and nothing new is bought on that check. The feed says so in plain words.
Stops keep running in code the whole time. The next check decides normally.
Your funds sit in your own Lighter spot account the whole time.
The agent can only place and cancel orders. Glint never takes custody, and you can pause an agent any time.
Every hour, each Glint agent writes a snapshot of its equity and PnL since start to Robinhood Chain (EpochRecorded).
The history is append-only. Nobody can edit it later, hide a bad week or restart the chart from a better day.
Glint pays the gas for these snapshots. That's why an agent's chart on the site is something you can check yourself, not a screenshot you have to trust.
Glint agents trade Lighter's spot markets on Robinhood Chain: tokenized stocks and ETFs like NVDA, TSLA, META, SPY and QQQ, all quoted in USDG.
For now it's spot only, with no leverage and no shorts. The agent can only sell what it actually holds.
New pairs are picked up automatically once they reach $5K of daily volume. When a market is closed, the agent can't buy it and just holds.
Next up: perpetual agents. They're in testing now and will come in a future update, with leverage limits enforced by the same risk engine.
Every trade a Glint agent makes gets a receipt on Robinhood Chain.
Side, market, size, price, the Lighter order reference. Each fill on the agent's Trades tab links to its receipt.
The record is public. Check it yourself.
What Glint costs:
- creating an agent: the gas for one transaction on Robinhood Chain. The key signature is free.
- each trade: no Glint fee
- AI model calls: paid by the platform
- trade receipts on-chain: paid by the platform
The only platform fee is on the marketplace. When your strategy sells, you keep 98% and the fee is 2%. The seller chooses whether to be paid in USDG or ETH.
What a Glint agent's model sees when it decides:
- prices and the indicators its strategy uses
- the strategy itself, in plain words
- its positions, as shares of the portfolio
What it never sees: your wallet address or your balance in dollars.
It isn't even called on every check. The agent only asks the AI when something changed: the price moved, a position got close to a stop, or the hourly heartbeat came due. A quiet market costs nothing and decides nothing.
The AI proposes a trade. Plain code checks it against your limits.
5% when the max is 10%: the order goes to Lighter.
12%: rejected, and the reason shows on the agent's page.
The model never has the final word on your money.
For everyone asking about how fees, buybacks and burns work on Glint, here’s a quick breakdown.
We’re building for the long term, and we want the platform’s growth to benefit both holders and builders.
Appreciate the support man!
All revenue from the platform goes back into buybacks, including marketplace sales and commissions.
We also locked 1% of the supply for the dev and bought back another 0.5% on top of that.
Part of the fees goes toward buybacks + burns, while the rest rewards people building AI agents on our platform.
Will think about DEX boost.
Appreciate the support man!
All revenue from the platform goes back into buybacks, including marketplace sales and commissions.
We also locked 1% of the supply for the dev and bought back another 0.5% on top of that.
Part of the fees goes toward buybacks + burns, while the rest rewards people building AI agents on our platform.
Will think about DEX boost.
Premium for $GLINT holders
- decision intervals from 1 minute instead of 5
- premium AI models
- a higher daily limit for chatting with agents
Premium is checked on-chain from your balance on every decision. If your balance drops below the minimum, the agent falls back to the standard interval and model.
Why we’re building Glint
We know everyone’s posting a thesis right now, so here’s ours. We think trading agents will create a new market for something traders have always valued but had a hard time buying: someone else’s way of trading. A trader who’s spent years refining a strategy should be able to turn it into an agent, show people how it trades, and earn when someone wants to run a copy. That’s why we’re building Glint.
Think about how trading ideas travel today. Someone posts that they’re buying ETH on the way down. You see the entry, maybe an update a few days later. What’s harder to pick up is how they sized that first buy, why they waited before buying again, or why they stopped while they still had cash. You can agree with the original call and end up trading it completely differently. A lot of a trader’s skill lives in those decisions.
What caught our attention when @Lighter_xyz brought its AI Agent Kit to Robinhood Chain was the ability to test a strategy against a live order book. You could watch an agent make those decisions as the market moved, paper trade it, and decide when you were comfortable putting money behind it. That got us thinking about both sides of the trade. What would make a trader confident enough to publish an agent? And what would another trader need to see before buying it?
On Glint, you start by describing your strategy in your own words. You choose which assets your agent can trade, how much it can spend, and how much it can hold. It can trade ETH, tokenized stocks, and ETFs on Lighter’s spot markets on Robinhood Chain. The agent proposes an order, and Glint checks it against your limits before sending it. If you’ve capped your ETH position, the agent has to respect that cap even when its next decision is to buy more.
The marketplace is the part we’re most bullish on. Before you buy an agent, you can read its full strategy and look through its onchain trading history. Say you’re comparing two ETH agents after a selloff. Both finished the week around the same place. One spent its budget on the first drop and sat through the rest. The other bought in smaller amounts over several days. Seeing that difference tells you something a return figure alone never could. You can decide which approach you’d actually want to run.
For the creator, that same history becomes part of what they’re selling. They can list the agent they’ve been running, and buyers can judge the work for themselves. Each time someone buys a copy, the creator earns from the sale.
Long term, our bet is that people will follow agent builders the way they follow traders today. If you’ve run someone’s ETH agent for six months, you have a pretty good reason to pay attention when they publish their next strategy. We think entire businesses will grow out of that relationship. We’re building Glint to be the place on Lighter where a trader can find their first customer and give them a reason to come back.
https://t.co/ZRdPlNsgaX
https://t.co/Fothp4v3NV
Lighter AI Agent Kit now supports Robinhood Chain!
Test your strategies against live order books, monitor their performance with the paper trading engine, and deploy them to production.
Compatible with your favorite harness.
Install in one line!
https://t.co/XM5A7Y4IuC
Five presets to start from:
DCA buys in portions.
Protect runs your stops.
Trend rides the EMA cross.
Rebalance holds target weights.
Momentum catches breakouts.
Or skip them and describe your own plan.
Glint ships an MCP server.
Your own AI assistant can draft, validate and prepare a Glint agent from one sentence, then watch or stop it.
It never signs. You sign on the site.
What's checked before an order
What every action has to pass before a Glint agent sends an order:
- size vs your max per trade and max per token
- daily loss: past your cap, the agent stops buying until the next UTC day. Sells and stops keep working.
- trades per day
- the market is in the agent's spec, enabled and open
- liquidity: at least $5K of 24h volume, a spread no wider than 1.5%, and an order no bigger than 2% of daily volume
- price: limit orders within 1% of mid, and the full size has to fit in the live book
- a 10 USDG minimum order
Your limits are written into the AgentHub contract when the agent is created.
Glint roadmap
Now: AI agents trading tokenized stocks on Lighter, plus a marketplace for AI agents
Next:
- rewards for agent creators
- perpetuals
- more markets beyond stocks
- a no-code builder for AI agents: put your agent together from blocks (signals, entries, exits, limits)
https://t.co/ZRdPlNrIlp
Write your plan in plain words.
"Buy NVDA on dips, once a day. Never more than 30% of the capital in NVDA."
Glint turns that into a strategy: markets, schedule, limits. Every field is validated before anything goes live.