Coding agents, by discoverability
Agents don’t live in one store with one search box — so the first question is simpler than “which is best”: which ones get found at all? This is the shelf as a buyer meets it today: the coding category of the AI Agents Directory, in its own order, measured October 21, 2026.
| # | Agent — as described in its listing | Directory score |
|---|---|---|
| #1 | a0.dev Directory badge: “Best overall” · Freemium · Closed source | — |
| #2 | Micro Agent Directory badge: “Best free option” · Free · Open source | — |
| #3 | Zugabot Directory badge: “Best API option” · Paid · API | — |
| #4 | bolt.diy AI-powered full-stack web development with multi-LLM support in your browser | 75 |
| #5 | Lovable The world’s first AI full-stack engineer that builds software 20x faster | 72 |
| #6 | Potpie AI Build custom AI agents for your codebase in minutes | 72 |
| #7 | Bolt AI-powered full-stack web development platform | 70 |
| #8 | Devin AI Autonomous AI software engineer developed by Cognition | 68 |
| #9 | Floot Vibe coding, no code — AI app builder, website builder | 63 |
| #10 | Devra Your AI coding worker bee | 60 |
| #11 | Replit Agent AI-powered coding assistant integrated into Replit’s browser-based IDE | 58 |
| #12 | v0 AI-powered UI generator for rapid prototyping and development | 58 |
Found is not the same as working
A directory standing says an agent is discoverable. It says nothing about whether it survives contact with a real support queue. That second axis — effectiveness — is being built the slow, honest way: structured reports from identified builders in the task rooms (what they run, for which task, for how long, and what it replaced). When both axes exist for a task, the interesting agents are the ones where they disagree: heavily discovered and barely vouched for is marketing; vouched for and barely discoverable is an opening.
Follow the build in the community →
How this page is measured
Standings are the order agents appear on the AI Agents Directory category page, read top to bottom on October 21, 2026. The first four slots carry that directory’s own editorial badges (best overall, best free, best open-source, best API); the score column is the directory’s published score out of 100, not a RankBlob metric. This page measures discoverability — how findable an agent is — not how well it works. Effectiveness standings will come from structured reports by builders in the task rooms, never from anonymous votes. One directory is a pilot sample, not the market; more sources join as the series expands.