Intro
A few interesting projects caught attention on Hacker News this week. Not the "big lab published a paper" kind of news, but actual software you can install and run today. Here are three open-source tools, each tackling a different pain point in the developer workflow.
KanBots: Every Kanban Card Gets Its Own AI Agent
If you've used a kanban board before, you know the drill: create a card, write a description, wait for someone to pick it up. KanBots swaps the "someone" for an AI agent. Each card can spawn a dedicated Claude Code or Codex agent in its own isolated git worktree.
It's not a plugin or a CLI wrapper — it's a standalone desktop app (macOS / Linux / Windows), MIT licensed. Drag a project folder in, it generates a board automatically. The AI agents figure out how to split tasks, assign work, and verify each other's output.
The Autopilot mode is particularly nice: define a few personas, and AI agents auto-assign tasks, run in parallel, then review each other's output. If you're the type who commits before bed and checks CI results in the morning, KanBots brings that same async workflow to feature development.
Zero telemetry. V1.0 just shipped on HN.
Superset: A Task Manager for AI Agents
Superset (YC P26) calls itself "the code editor for the AI agents era." In practice, it lets you run 10+ coding agents simultaneously on your machine, each isolated in its own git worktree. You monitor progress, inspect diffs, and hop into workspaces from one interface.
Supported agents include Claude Code, Codex, Cursor Agent, Gemini CLI, GitHub Copilot, OpenCode — basically anything that runs in a terminal. 11k+ stars on GitHub, nearly 3,000 commits.
For developers who regularly have AI writing code, refactoring, or fixing bugs across multiple tasks, Superset solves a real problem: keeping your sanity when 5 agents are running at once.
Currently optimized for macOS. Windows/Linux support is in progress.
Models.dev: AI Model Specs, Open Source Style
Models.dev is an open-source database of AI model specifications. It covers 350+ models across chat, image, and video categories — specs, pricing, context lengths, benchmark data — all organized as TOML files in a GitHub repository.
What sets it apart: the data is publicly maintained via PRs, and there's a curl-friendly api.json endpoint. If you frequently compare model pricing, context windows, or capabilities across providers, this is much faster than digging through individual pricing pages.
Currently at 4k+ stars, 986 forks, maintained by the OpenCode team.
Wrap-up
All three are open-source or developer-friendly, each solving a distinct workflow problem. KanBots brings AI automation to kanban management. Superset solves multi-agent orchestration. Models.dev fills the information gap around AI model specs. Worth a look if you're optimizing your AI development stack.
