---
title: "mrin — about"
description: "mrin is the execution layer for AI agents: Firecracker microVMs where every command is a commit you can fork, restore or replay. Private alpha, built by a small team in Bengaluru, India."
url: https://mrin.ai/about
---
Company

# About mrin

mrin is the execution layer for your agents. Each run gets a real Firecracker microVM on Linux + KVM, and every command the agent runs is a commit of that whole machine, memory and disk. Any commit can be forked into a second machine, restored as a new branch, or read back exactly as it printed.

## Who it is for

Engineers at teams that already run coding, search, RL or eval agents, for whom fork and rollback sit on the hot path: coding-agent companies, RL-environment vendors, and eval harnesses. When a run fails at step 40, the usual answer is to rebuild the environment from zero and hope the failure reproduces. mrin lets you go back to step 39 instead.

## What we build

- A control plane written in Rust, one process with Postgres behind it, where the branch state machine is a pure function.
- A CLI and SDKs in TypeScript, Python and Go, built from one contract and checked by one shared conformance suite.
- Fork, restore, checkpoint, suspend, hibernate and scale on a running microVM, plus one WebSocket that streams every branch a client watches.

Version 0.1.0 was released on September 14, 2026.

## Where it stands

mrin is in private alpha. Today it runs on your own Linux + KVM host and we help you deploy it. A hosted version is in progress. The home page lists, row by row, what is not available yet.

## Where we are

We are a small team based in Bengaluru, Karnataka, India. Write to amit@mrin.ai, or [schedule a 30-minute call](https://cal.com/boomage-engineering/30min) and bring a workload. The [docs](https://docs.mrin.ai) cover the API, the CLI and the SDKs.