Anmol Sharma

Like to build stuff !

[Anmol Sharma @JavAnmol]

https://t.co/nSM4u4AV4m

[Anmol Sharma @JavAnmol]

One more open in Vitess 👀 Reserved gb18030 sequences are valid characters, not malformed bytes — Convert now quietly yields `?` instead of failing, matching MySQL 8.0.46.

[Anmol Sharma @JavAnmol]

https://t.co/hUyhOBvcTS

[Anmol Sharma @JavAnmol]

First PR to @vitessio 🚀

DO statements were being shipped raw to one arbitrary shard, quietly breaking locking functions and subqueries. This gives DO its own plan node and routes it like the equivalent SELECT.

[Anmol Sharma @JavAnmol]

Github Link: https://t.co/aS4IR4qfhi

[Anmol Sharma @JavAnmol]

Built a mini Vitess for Postgres - a lightweight sharding proxy that transparently routes queries across DB shards.

[Anmol Sharma @JavAnmol]

I built kira-mcp. a local MCP server that gives any AI agent (Claude, Cursor, Cline) computer-use: it sees your screen and drives mouse + keyboard. Local OmniParser vision, no API key, weights ship in the wheel.

pip install kira-mcp
https://t.co/VvxxHcY8ND

[Anmol Sharma @JavAnmol]

Trying to connect with more founders and builders in the AI and SaaS space.

Feels like everyone is building something interesting right now and I genuinely enjoy seeing what people are working on.

If you're building in a similar space, feel free to follow and connect.

[Anmol Sharma @JavAnmol]

I think we broke something.
Watch an AI agent solve a CAPTCHA like a human in real-time.

📦 PyPI: https://t.co/8LkzT1wnrU
⭐ GitHub: https://t.co/Etf0EvVi1c

#ai #claude #gemini

[Anmol Sharma @JavAnmol]

I built an AI agent that hires its own team.
One orchestrator. Automatic role assignment. Configurable depth and breadth.
Runs in your terminal. BYOK.

Live at: https://t.co/EZ0GdcoKrd
npm install -g axon-cli
https://t.co/FqBXcwS3Xc

[Anmol Sharma @JavAnmol]

ORAG is live. GitHub link : https://t.co/MwP4iStK2e
Built this in TypeScript - LangChain.js for RAG, MCP for the agent interface.
If you're building AI systems on top of internal data, this is the missing context layer. Would love feedback from anyone working in this space.

[Anmol Sharma @JavAnmol]

I built ORAG - an organizational RAG + MCP platform in TypeScript.

Live at https://t.co/pPNLl4L9CJ

Feed it your org's docs, wikis, and databases. It exposes them as a fully typed MCP server your AI agents can call.
Here's what I built and why. 🧵

[Anmol Sharma @JavAnmol]

So I built a platform that solves exactly that.
ORAG does 3 things:

→ Indexes your internal docs into a RAG pipeline (chunking, embedding, retrieval)

→ Wraps that retrieval layer as an MCP server

→ Adds org-grade access control so the right agents see the right data

[Anmol Sharma @JavAnmol]

Drop one AI tool you actually use daily - no hype, just real workflow.

I’ll start: Perplexity AI for quick documentation lookups.

[Anmol Sharma @JavAnmol]

I wrote a full breakdown , a comparison table, and when each fails you - link in bio.
If you're building AI features in TypeScript, follow me - I write about this stuff every week.

[Anmol Sharma @JavAnmol]

Vercel AI SDK vs LangChain.js.

I've used both. Here's when to use which and when each one will waste your time.

[Anmol Sharma @JavAnmol]

The honest recommendation for most developers:
- Start with Vercel AI SDK.
- Get to production. Ship the feature. Understand the problem space.
- Then reach for LangChain when you hit something it genuinely solves better.
- Don't use both in the same project without need.