I build AI products that ship, and the systems that keep them honest.
AI agents, permission-aware RAG, voice and agent evals, and the full-stack around all of it, plus an offline mesh messenger for when there is no internet. Fifteen-plus products and tools live, shipped in days. Open source, every single day.
My 2026, in green.
3,130 contributions in 2026 alone. GSoC, the Linux Foundation (LFX), and Symmetry Autumn of Code. The graph is the habit, not a sprint.
github.com/royalpinto007 ↗
Real products, real users.
Not portfolio mockups. Each one is deployed, maintained, and one click away.
SignalizeAI↗
Turns any company website into a sales-ready prospect record: strategy, personas, outreach. Browser extension + web app + billing, running in production.
AashiNyra Deals↗
Live women's fashion deals platform: tracks 16+ retailers nightly for genuine price drops, with cashback to UPI. Next.js on Cloudflare Workers, Supabase, D1/R2.
Kabootar↗
Offline mesh messenger: chat with no internet, no servers, no SIM. Messages hop phone to phone over Bluetooth and Wi-Fi, a delay-tolerant network with epidemic routing and end-to-end delivery acks.
Tiny Day↗
A cozy, offline-first daily planner for Android: gentle scheduling, focus, and replanning, with no accounts and data kept on-device.
Auralis↗
A living field of light rendered in real time on the GPU, in the browser: WebGPU with a WebGL2 fallback, no server. Move your cursor and the current bends around it.
Mycelia↗
A GPU slime-mold simulation in the browser: 262k agents leave trails, sense them, and self-organize into living networks that flow toward your cursor. WebGL2 GPGPU, no server.
Ten sites, one bar.
Design-forward, multi-page marketing sites, each a distinct genre and visual identity, built static-first and verified to render fully without JavaScript. Need a landing page that looks like this? That is the point.
Five agents that do things.
Built to prove one point: agents should take real, auditable actions, not just draft text. Try any of them, no signup.
Want the source of all five? It's $49 as a starter template.
The systems underneath the agents.
I pick the right substrate for your stack, not the one I happen to know. These are live, and each links to open source.
RelayG↗
My production triage agent rebuilt as a LangGraph state machine: typed state, SQLite checkpointing, and a human-in-the-loop interrupt that pauses the graph for refund approvals.
Casebook MCP↗
A live Model Context Protocol server that makes the AgentPostmortem failure registry queryable by any agent (Claude, Cursor, anything). Four tools, streamable-HTTP transport, plus a Claude Agent SDK investigator.
Casebook Chat↗
A streaming investigation assistant on the Vercel AI SDK: the model tool-calls the MCP server above and cites real cases, with tool invocations rendered inline as it thinks.
VaultRAG↗
Permission-aware RAG where the access-control check lives inside the retrieval query. A gold-set eval fails CI on any leak: delete the predicate and leak rate goes 0% to 81.8% while recall stays 100%.
agentrace↗
Observability for AI agents. Parses Claude Code session transcripts with no instrumentation and flags results you should not trust, every check drawn from a real failure across 152 production runs.
voiceeval↗
Evaluation for voice agents. Timed turns with a truth field per case, catching misheard numbers (fifteen vs fifty), missing confirmations, and blown latency budgets that text evals miss.
mcp-audit↗
A security scanner for MCP servers, like npm audit but for the Model Context Protocol. Connects over stdio or HTTP, runs 18 rules, and emits SARIF straight into GitHub code scanning.
evalgate↗
Prompt and agent regression CI: the build fails when your prompt gets dumber. Ten scorers, a baseline and delta engine, and a GitHub Action that comments the quality delta on every PR.
tokencut↗
Measure and cut the token cost of an LLM or agent message payload before you send it: truncate bloated tool results, drop duplicate context, and trim history to a budget. Zero model calls, deterministic.
answerproof↗
Verifiable, tamper-evident receipts for RAG answers. Records which sources were retrieved and cited, with a Merkle root and an Ed25519 signature anyone can verify independently.
ctxlens↗
A context-window profiler for AI agents. Parses session transcripts, shows what is eating your context per turn, flags duplicated and stale content, and recommends what to cut.
injection-arena↗
A self-hostable prompt-injection challenge game. Ten levels stack real defenses, a server-side judge scores every crack, and the leaderboard runs live on Cloudflare Workers and D1.
distro-skills↗
26 Agent Skills that teach an AI agent to distribute a developer product: Show HN, Product Hunt, awesome-lists, dev.to, GitHub, Reddit, X, npm/PyPI. The builder's counterpart to marketing skills, honest and channel-specific.
Pick a scope, get a working result.
Scoped Agent Build
- A production agent or AI feature with real tool-use, approvals, and guardrails
- On your stack: n8n, Claude API, Next.js, Supabase
- Deployed + documented, Loom walkthrough, handover doc
- 14 days of fixes after delivery
AI Automation Sprint
- Discovery call to pick the highest-pain workflow
- Working n8n + Claude automation, end to end
- Loom walkthrough + handover doc
- 7 days of fixes after delivery
Automation Retainer
- Ongoing builds + maintenance
- Priority turnaround
- Monthly review of new automation opportunities
- $150 automation audit credited if you start with one
Or ship your own agents this week.
Agent Suite Starter
- Full source for Resolvd, Tracecase, Webhands, Bridgekit, Greenlite
- The direct Workers-to-Groq AI pattern
- Per-app env examples + Supabase schemas
- Deploys entirely on Cloudflare + Supabase free tiers
Tell me the task you hate doing.
I'll reply with whether I can automate it and what it would take. No pressure, no spam.
Prefer email? royalpinto007@gmail.com · I reply within a day.