what parasol is, and why i built it

parasol is the trading agent that protects you from solana. it scans every new token, filters out the scams, and trades the rest with mpc-secured wallets, ai-managed exits, and a safety layer that any other ai agent on the network can call before it buys anything.

that's what it is now. this post is the longer version — what i was doing before this, the five products i built that didn't quite become parasol, the month we just had, and the year i'm betting the rest of my work on it.

---

where i came from

my first real job in crypto was at magic eden. magic eden taught me that the difference between a marketplace people use and one they don't is almost never about the technology — it's about whether the product respects the user's time, money and attention. you can ship the most clever indexer in the world and lose to a slower competitor with a cleaner ui. the bar is set by the people who already trust you.

after magic eden i went to liquidloot. liquidloot taught me a different lesson: that on-chain financial products live and die by execution reliability, not by the elegance of the model behind them. a 90% success rate on swaps is not a 10% problem. it is a complete trust failure. the users you lose to a failed trade do not come back to read your post-mortem.

between then and now i shipped five products of my own. some of them did fine. some of them did not. all of them taught me what to delete next time.

what they had in common was that i was building features and waiting for distribution. what changed in 2026 was that i stopped doing that. parasol is the first product where i started with the assumption that the engine had to be obviously real, the safety had to be visibly enforced, and the people i sold to had to be able to verify everything i told them by looking at the code. everything else followed from that.

---

the bet

solana ships ten thousand four hundred new tokens a day. ninety-eight percent of them die. the two percent that don't move faster than any human can react.

three things follow from that:

parasol is the bet that an autonomous ai agent — one that scans every source, filters the manipulation, scores what survives, executes through non-custodial wallets, and manages every exit — can do for a normal solana trader what a hand-built bot does for an insider.

and that the same intelligence layer, exposed as a hosted api with metered billing, becomes infrastructure other builders can plug into.

---

what parasol is, in detail

we have three products that live on the same engine.

under the hood: a 6-layer manipulation filter, multiple proven trading personas (sienna, claudia, yuki, ivy, grok, ghost, scalper), a scoring engine that improves with every closed trade, kill switches and concept-drift detection on every account, mpc-secured non-custodial wallets, and an mcp server that any ai agent — claude, cursor, the new phantom mcp, custom builds — can call to check whether a token is safe before it buys.

the safety layer is the part we're proudest of. one tool call — parasol_safe_trade_check — runs rug detection, honeypot simulation, holder analysis, liquidity adequacy, and slippage estimation. it returns a clear verdict: safe / caution / avoid, with the specific risk flags. an ai agent that calls this before any swap stops being a hands-without-eyes liability and starts being something a wallet team can trust.

---

what we shipped this month

the last two weeks have been the most relentless engineering window in parasol's history. four structural transitions in parallel. each gets its own post in this series:

we're publishing each of those as a standalone post over the next ten days. a cycle-style hackathon recap at the end of the window will tie them together with the numbers.

---

why solo, why now

i shipped five products before this one. the lesson from each of them was that distribution is the hardest part and that nobody is going to build it for you. parasol is the first product where i went all in on a single thesis — that ai agents and on-chain finance need to fuse, and that the trading agent is the cleanest demonstration of why — without leaving myself an exit.

the team is small. solo, by founder count, although we work with a network of contributors who've shipped real code. cap, who runs colosseum, has written that solo founders win when founder-market-fit is clear. mine is.

magic eden gave me retail crypto distribution. liquidloot gave me on-chain trading reliability. five products of my own taught me what to delete. parasol is what's left.

---

what i want from you

i want feedback from people who actually trade on solana. not survey feedback. real feedback from someone who's lost real sol and wants to lose less of it.

parasol.so/beta captures early access requests. dms to @parasolsolana get answered, mostly by me.

if you're a builder, the sdk is mit. build something with it. tell us what's missing.

if you're a researcher, the whitepaper is public and the open-source repo includes the documentation, types and example client.

if you're a wallet team — phantom, jupiter, or anyone else thinking about how ai agents should interact with onchain execution — the mcp endpoint is live and we're already running an integration guide for parasol + phantom mcp that any of your users can set up in five minutes.

---

the next year

the rest of the year sits in three buckets:

we'll write posts on each of those when they ship. we won't promise anything that isn't either already in the codebase or two weeks from being there.

---

we built the product i've wanted on every chain for the last five years. it sees what you can't. it stays awake when you sleep. it stops you from buying the bad ones. and when one of the good ones runs, it manages the exit so you keep what you made.

if you've made it to the end of this post you're the kind of person we built parasol for.

— the parasol team