How We Rank AI Agents
Each product on a category page gets a score out of 100, built from six parts: fit for the job (30), autonomy (20), controls (15), integrations (15), access (10) and evidence (10). Rank is sorted by total, then by fit for the job, then by name. A row is marked partial fit when job fit is under 18. Facts come from the vendor's own public pages, checked and dated; this edition was checked on 27 October 2026. Prices are shown in USD. Where a vendor page only shows euros or pounds, we record the price as shown and say so. If a vendor doesn't publish a price, the listing says 'not published'. We do not print a third-party guess.
1 · 30 points
Fit for the job
Does this product do the job in the category, or is it only adjacent to it?
2 · 20 points
Autonomy
Copilot means a person drives every step. Semi-autonomous means the product can run a stretch of the work and stop for approval. Autonomous means it is sold to operate the motion.
3 · 15 points
Controls
Can a team review, edit, approve, and stop the work before it reaches a customer or a creator?
4 · 15 points
Integrations
Does it connect to the systems the team already uses?
5 · 10 points
Access
Is there a free tier or a published price a team can read before a sales call?
6 · 10 points
Evidence
Is the product public enough to describe without guessing: a live site, a pricing page or an honest absence of one, and a clear description of the workflow?