YEN
JUTSU
ONE GUY.
ZERO BUDGET.
FULL SWARM.
Anuj runs M4ST solo out of Bareilly, India — a squad of AI agents built with no funding, no team, and no cloud bill, trained to plan, patch, and ship on their own.
Anuj runs M4ST solo out of Bareilly, India — a squad of AI agents built with no funding, no team, and no cloud bill, trained to plan, patch, and ship on their own.
Every arc needs a starting point. Here's his.
No mentor arc, no funding montage — just a laptop and the stubborn idea that a smart system shouldn't need a data center behind it.
Every piece of M4ST got built the hard way: self-taught, self-funded, self-hosted. While everyone else rented GPUs, he stitched together eleven free-tier providers into one routing chain and kept inference sitting at zero.
Currently on a side-quest toward a CEH certification, taking freelance AI and automation work on the side. Jugaad rule: shipped beats planned, every single time.
What actually goes into a build like this.
Swarms that split work and delegate — nobody's watching the loop.
21 custom protocol bridges linking models to real tools.
11 free providers, one failover chain, ₹0 spend.
Quantized models running on a consumer GPU, not a rented one.
12ms cache hits, 40% fewer repeat calls.
CEH-track recon work, applied to real audits with real sign-off.
Live builds, not a pitch deck.
The full operator system — five agents, MCP orchestration, semantic caching, zero-cost routing, all running together.
A coding workspace wired straight into GitHub and Vercel — a merge lands, a deploy follows, no one has to push the button.
An open-source recon framework — Shodan data in, reasoned analysis out, no manual digging.
Sub-12ms lookups over embedded memory. The same question never gets asked twice.
How the build actually escalated, in order.
Free-tier routing across 11 providers finally held under real traffic — the ₹0 bet started paying off.
First custom protocol server shipped, then multiplied to 21 — databases, browsers, recon, all wired in.
M4STCLAW's first working multi-agent build went live, then kept getting rebuilt through v5, v6, v7.
Full provider stack corrected, smarter routing added — the arc isn't finished, it's just the current chapter.
Open to freelance AI and automation work, security consulting, and anything that needs someone who ships instead of stalling.