Vietnamese startup gives AI coding agents a memory boost

This article is a part of Startup Spotlight, a series that features young, up-and-coming startups.

Image credit: Timmy Loen

When Andy Nguyen, founder of Byterover, tried agentic coding tools, every lesson vanished between sessions. No shared memory, no reusable context. That irritation became a product focused on one question: how should AI remember code work?

😟 Problem

Developer teams now rely on AI coding tools such as Cursor, Claude Code, and GitHub Copilot. These tools generate large portions of code, yet forget past fixes, project rules, or edge cases.

Teams must repeatedly re-teach agents and manually paste specs or prior conversations. Context becomes scattered across chats, docs, and tickets, which raises the risk of wrong suggestions and slows down delivery.

As AI coding scales to tens of millions of users and a significant share of big tech code, the lack of persistent, sharable memory turns into an infrastructure gap.

💡 Solution

Byterover builds an agentic memory layer that plugs into coding agents through the Model Context Protocol, or MCP. It acts as a dedicated context engineer for AI inside integrated development environments (IDEs) and command-line interfaces.

• Auto-captures developer and LLM interactions into a structured memory workspace
• Lets agents retrieve precise memories for each new coding task
• Supports team workspaces where developers share and reuse AI memories
• Offers Git-like actions on memories, including create, update, rollback, and conflict resolution
• Positions itself as “Git for AI memory” for coding-focused agents

Image credit: Byterover

📊 Market size

The firm estimates a current AI coding market of about US$5.2 billion, using roughly 22 million AI IDE accounts and a monthly price near US$20.

They extend this view to the wider pool of 47.2 million software engineers worldwide, plus adjacent roles such as product managers and founders. With enterprise pricing and more agents, they forecast a market that could exceed US$20 billion.

🤝 Team

• Andy Nguyen, founder and CEO, former founding engineer at Cinnamon AI Labs, AI PhD dropout from the University of Victoria in Canada
• Ha My Tran, growth lead, previously grew a multi-agent workflow app to more than 1,000 global paid users in five months
• Engineering team in Vietnam with experience at VinBigData, FPT AI, and the AI department at UIT

🚀 Traction

• Shipped weekly for two months, reaching product version 2.0 and open-sourcing the core repo
• Daily paid users from multiple markets, with concentration in the US and Japan
• Recorded a team subscription worth US$1,780 per year
• Inbound enterprise interest, including leads from Fortune 500 companies
• More than 3,000 GitHub stars on the open-source repository

🏆 Competition

💰 Financials