AutoBot

A self-improving agent harness for long, complex knowledge work

🤖 AI Agents ⚡ ProductivityOpen SourceY▲ 20 on HNLaunched October 2, 2026

AutoBot is a self-improving agentic harness that helps frontier AI models finish complex knowledge work. It repairs and validates its own harness so later workflows inherit improvements, keeps hierarchical memory on disk, and uses persistent task graphs and checkpoints so work can resume if an agent stops. First shared by its maker on Show HN: https://news.ycombinator.com/item?id=49743478

✨ The overview, FAQ and details below were drafted by AI from the product's own website and haven't been reviewed by the maker yet. Are you the maker? Claim this listing to review and edit them.

Overview

Who is it for?

People who delegate long, multi-application knowledge work to AI agents.

Problem

Agents lose context and progress on long tasks, and each hard workflow starts from scratch.

Solution

A harness that improves itself from hard workflows, builds task-specific context from on-disk memory, and checkpoints progress.

What makes it unique

The project reports 18.5% higher task completion than a published OpenAI baseline on OSWorld 2.0 and #1 on the AssistantBench hidden-test leaderboard.