How I built a real-time pump.fun deployer tracker and connected it to KOL wallet monitoring on Solana
I run MadeOnSol, a Solana ecosystem directory. Over the past few months I built two intelligence tools on top of it — Deployer Hunter and KOL Tracker — and recently connected them so they feed each other signals in real time.
This post covers the architecture, the data pipeline, and what I learned building it as a solo developer.
The problem
On Solana, thousands of tokens launch daily on pump.fun. Most are worthless. A few graduate to real liquidity.
The question every trader asks is:
"Did a good deployer just launch something, and are smart wallets buying it?"
Before I built this, answering that question meant manually checking deployer wallets, cross-referencing with KOL wallets, and hoping you caught the signal in time. I wanted to automate the entire pipeline.
Deployer Hunter — tracking 7,197+ deployers
The core idea is simple: track every pump.fun deployer, score them by their graduation rate (how many of their tokens actually bond to Raydium), and surface the ones with real track records.
Data ingestion
I use a gRPC stream connected to Solana via a Kaldera RPC endpoint. This gives me real-time transaction data as it hits the chain — no polling, no delays. Every time a token is created on pump.fun, I capture the deployer address, token mint, and timestamp.
Scoring logic
Each deployer gets two scores:
- Recent rate — graduation % of their last N tokens
- Lifetime rate — graduation % across all their tokens
Deployers are classified into tiers:
| Tier | Meaning |
|---|---|
| 🏆 Elite | Consistently high graduation rate across many deploys |
| ✅ Good | Solid recent track record |
| 🌟 Rising | New but showing promise |
Storage
Everything goes into Supabase (PostgreSQL). Deployer profiles, token history, graduation events, and peak market caps are all stored and queryable. Each deployer gets their own profile page with full history.
The feed
Deployer Hunter shows a live feed of new deployments from tracked deployers, filterable by tier. When an Elite deployer launches something, it shows up immediately.
Currently tracking 6,366 deployers with 194 bonds today at a 6.3% graduation rate.
KOL Tracker — 462 wallets monitored
KOL Tracker watches what the biggest Solana wallets are actually buying on-chain. Not what they tweet — what they trade.
Wallet monitoring
I maintain a curated list of 462 KOL wallets. The system monitors their transactions in real time using the same gRPC infrastructure. When a tracked wallet buys or sells a token, it appears in the live feed within seconds.
Features
- Live feed of all KOL buys and sells
- PnL leaderboard — who's actually profitable
- Top buys — which tokens are getting the most KOL attention
- Hot tokens — tokens with multiple KOL buyers
- Compare tool to put KOLs side by side
Users can suggest new KOLs to track. I verify the wallet before adding it.
The connection — where it gets interesting
The real power came from connecting these two systems. Now:
- When a tracked deployer launches a token → Deployer Hunter shows it
- When KOLs start buying that same token → KOL Tracker shows it
- Both feeds cross-reference each other — you see KOL interest on a deployer's launch, and deployer quality on a KOL's buy
Every KOL profile now has a Deployer Score — what percentage of their buys come from tokens launched by tracked deployers. This tells you whether a KOL is buying from proven deployers or chasing random launches.
The signal chain becomes:
Good deployer launches → KOLs buy → you see it on both feeds simultaneously.
Tech stack
| Component | Tool |
|---|---|
| Frontend | Next.js (App Router, SSG where possible) |
| Database | Supabase / PostgreSQL |
| Real-time data | gRPC stream via Kaldera RPC |
| Hosting | Hetzner VPS (CX23) |
| Resend | |
| DNS | Cloudflare |
Total infrastructure cost: ~€140/month.
What I learned
gRPC > polling
The difference between polling an RPC endpoint every few seconds and having a real-time gRPC stream is massive. You catch deploys and buys within the same slot they happen.
Scoring is harder than detection
Detecting a new deploy is easy. Deciding whether a deployer is "good" requires careful threshold tuning.
Early on, tokens were showing as "bonded" with peak market caps of $3-5K — impossible since bonding requires ~$69K. Turned out to be false positives in the bonding detection logic. Still refining this.
Kill features that dilute focus
I previously had five intelligence tools. I removed three (Wallet X-Ray, CTO Radar, Token Scanner) to focus on Deployer Hunter + KOL Tracker. The product is stronger for it.
The data is the moat
Anybody can build a UI. The value is in 7,197+ deployer profiles with historical data and 462 tracked KOL wallets with real PnL. That dataset grows every day and gets harder to replicate.
What's next
- Publishing monthly data reports using this pipeline (graduation rates, KOL accuracy, deployer trends)
- Improving bonding detection to eliminate false positives
- Building a TikTok-to-Token analyzer that correlates viral TikTok content with token launches
Both tools are free and require no registration to browse:
Happy to answer any questions about the architecture or the Solana data pipeline.
Top comments (0)