AmyAI
A self-hosted coding agent, fine-tuned on your own code. The aim is honest answers grounded in a real execution ledger, with a learning loop that improves over time.
In active developmentKey features
Self-hosted
Runs entirely on your machine. Your code never leaves your computer. No cloud dependency, no vendor lock-in.
Fine-tuned
Learns from your codebase. Understands your patterns, conventions, and the real shape of your projects.
Honest
An anti-fabrication gate is meant to catch false claims, so answers stay grounded in what the system actually did rather than what it wishes it did.
Ledger-backed
Real execution ledger. Every tool call is recorded. You can verify what the agent actually did, audit runs, and learn from failures.
Learning loop
Improves over time through lessons. Systematic learning from real failures and corrections, built into context, not weights.
Multi-agent
Specialized agents for different tasks. Builder, Coder, Explorer, Reviewer, Planner. Roster-based orchestration.
What sets AmyAI apart
Most AI coding assistants claim to help you write code. In practice, they guess at what you need and confidently produce the wrong thing.
AmyAI is built around honesty. The intent is for every answer to be backed by an execution ledger: if it didn't do something, it says so, and if a tool call failed, you see it. The learning loop is designed to feed on real failures, not hallucinations.
It is also fine-tuned on your own projects, so it can pick up your patterns and conventions instead of guessing at them.
Where it stands
AmyAI is in active development. It is one of the tools we are building in-house, not a product on sale today. This page is a look at what we are working on and the principles behind it.
It runs self-hosted, so your code stays on your machine and you can see exactly what the agent does on every run. We will share more as it matures.