September 2026 edition
An independent register of AI agents and harnesses. Each entry is rated from dated public evidence: adoption and release data, 286 independent reviews and 994 coded community posts. No sponsored placements; thin evidence is marked, not guessed.
Highest rated this edition
Stably AI · Orchestrators & meta-harnesses
Open-source desktop 'agent development environment' that runs many CLI coding agents side by side, each in its own git worktree.
Highest-rated entries across every category. Only medium- and high-confidence ratings are ranked; provisional ones follow. Sort any column; tick up to four to compare.
| No. | Compare | ||
|---|---|---|---|
| 1 | OrcaStably AI · Orchestrators & meta-harnesses | 78medium | |
| 2 |  OpenCodeAnomaly · Coding agents | 75high | |
| 3 |  Claude CodeAnthropic · Coding agents | 73high | |
| 4 |  Zed (Agent Panel)Zed Industries · Coding agents | 73high | |
| 5 |  GitHub CopilotGitHub (Microsoft) · Coding agents | 72high | |
| 6 |  Codex (CLI and app)OpenAI · Coding agents | 71high | |
| 7 |  SupersetSuperset Inc. · Orchestrators & meta-harnesses | 70high | |
| 8 |  Hermes AgentNous Research · Personal & desktop agents | 69high | |
| 9 |  Grok BuildSpaceXAI (formerly xAI) · Coding agents | 68high | |
| 10 |  HerdrHerdr (Can Celik; YC F26) · Orchestrators & meta-harnesses | 68high |
Where independent reviewers disagree with what users actually say. The gap is often the most useful thing to know before you choose.
The top-ranked entry in each category.
Coding agents
Anomaly
Orchestrators & meta-harnesses
Stably AI
Personal & desktop agents
Nous Research
Browser & computer use
Browser Use (browser-use.com)
Agent-building platforms
n8n GmbH
Business & sales agents
Clay Labs
Entries renamed, transferred or struck off since the April 2026 edition, each with its source on the entry.
Gemini CLIGoogle announced on 2026-05-19 that it is 'transitioning our terminal experience from Gemini CLI to the brand-new Antigravity CLI'. From 2026-06-18 Gemini CLI stopped serving requests for free-of-charge users and Google AI Pro/Ultra subscribers (also Gemini Code Assist IDE extensions and Code Assist on GitHub free usage).
Vibe KanbanBloop, the company behind Vibe Kanban, shut down on 10 April 2026. Cloud/remote features (kanban issues, comments, projects, organisations) were removed 30 days later; local workspaces keep working and the Apache-2.0 code is 'community maintained'.
In maintenance mode: a project collaborator wrote on 2026-05-01 that Gas Town is 'staying focused on infrastructure and reliability fixes only' and pointed users to its successor, Gas City. Last release v1.2.1 on 2026-06-06; last commit on main 2026-07-23.
Kilo CodeAnaconda announced its acquisition of Kilo Code on 15 July 2026; the product continues under the Kilo brand with no announced changes to products or plans ('Kilo remains available today with no changes to existing products, plans, or support for current users').
ContinueAcquired by Cursor around 16 June 2026 in what press describes as an acqui-hire; the standalone product was discontinued. The continuedev/continue repository is read-only and 'no longer actively maintained'; a final 2.0.0 release (VS Code extension, CLI, JetBrains) shipped 19 June 2026 with telemetry and authentication removed.
Deprecated in February 2026 and replaced by Nimbalyst (nimbalyst.com, github.com/nimbalyst/nimbalyst). The final Crystal release v0.3.5 (2026-02-26) says it 'will not be updated in the future'.
Fin (formerly Intercom Fin)Intercom renamed itself Fin on 2026-05-12; Salesforce announced a ~$3.6B acquisition on 2026-06-15 and completed it on 2026-09-10. Fin continues to be sold, including on non-Intercom helpdesks, as part of Salesforce AI Labs.
FlowiseWorkday acquired Flowise on 2025-08-14. On 2026-07-29 the team announced a wind-down: code freeze 2026-07-29, GitHub repository archived 2026-08-13, end of life 2026-08-31.
Three pillars, public weights, and a minimum of evidence for each. A pillar without enough evidence is left out rather than guessed; an entry needs two pillars to be rated.
Stars, downloads, installs and reported users on a log scale, plus releases in the last 90 days.
Verdicts from named independent reviewers in the last 12 months, averaged toward neutral when there are few. Benchmark placings are shown on every entry but not scored: they mostly measure the model.
A coded sample of recent Reddit, Hacker News, X, YouTube and GitHub posts. The sample size is always shown.
