Gujarat, India · Since 2025
Tagbudy Services builds developer tools on top of Claude. Its first product, BEHAVR, gives AI apps a way to remember what they see.
Built and maintained by the founder. Shipped publicly, with an SDK on npm and an MCP server for Claude Code.
A persistent visual memory API. Send in an image, screenshot, document or video. Ask questions across everything an app has stored. Two calls: remember() and ask().
ILLUSTRATIVE FLOW
How BEHAVR works
Step 1
An image, screenshot, document or video goes in through one API call.
Step 2
Claude's vision turns it into structured memory that stays searchable for that user.
Step 3
Questions run across everything stored, and answers point back to where they came from.
Use cases
Any product whose users send images, and which should remember them later, can use BEHAVR. These are the kinds of products I am building it for.
A robot that remembers what it has seen: where things were placed, which aisle was blocked, what changed in a room.
Home, warehouse, service and hobby robot builders.
A record of what the wearer saw during the day, searchable in plain language, including for users with low vision.
AI glasses, lifelog and accessibility apps.
Snapshots turned into a history the app can answer questions about, instead of a folder nobody scrolls through.
Home camera, small shop and fleet apps.
Bills, warranty cards, scanned papers, notes and trip photos become something users can ask questions about.
Photo organizers, expense, home inventory, scanner and notes apps.
Sites where users already upload photos, and the site should remember them later, such as property listings, damage claims and repair requests.
Real estate, insurance, repair and e-commerce sites.
Assistants that take images as input and should still remember them next week: a chart, a homework page, a meal photo.
Personal assistant, study and non-clinical fitness apps.
Claude Code, Cursor and Claude Desktop users who share screenshots and UI bugs can keep them as memory through the MCP server.
Developers, QA and design review tools.
EXAMPLE QUESTIONS ARE ILLUSTRATIVE. THESE ARE SEGMENTS I AM TARGETING, NOT A LIST OF CUSTOMERS.
Journey
I have built and shipped several products on my own. Many did not take off, and I am open about why. Each one taught me something, and I carried it into the next.
A persistent visual memory API for developers. I repositioned it several times as I learned what developers actually need. The gap today is reach: getting it in front of enough developers.
A command-line tool that extracts a developer's coding style for Claude Code, published on npm. I paused it because users could export their style and use it elsewhere, so there was little reason to stay.
A cold-email platform for newly registered domains. Four backend services built, then shelved: little demand pull and a deliverability risk.
A multi-AI debate platform with ChatGPT, Claude, Gemini and Grok, built solo in about ten days. It shipped with little exposure. Distribution was the gap.
A platform where an idea got an AI-based value, and that value let its owner set up a share price, like a stock market. It got little exposure, and I pivoted before testing it in the market. The lesson: put the product in front of people first, then decide.
A prompt-to-build-your-own-AI tool for any industry. It got little exposure and I pivoted before testing it in the market. The lesson: put the product in front of people first, then decide.
An email outreach product. I built and shipped it, but it did not get enough exposure or distribution.
Built, failed, learned, built again. That is the pattern, and I am not stopping.
Tagbudy is a solo-founder company based in Gujarat, run by Bhavin Patel. It builds AI software for developers, and Anthropic's Claude models sit at the core of its product.
It is early stage. I ship in public, publish SDKs and open-source tools, and work directly with the developers who use them.
Contact