PickMyAgent

Methodology 1.0

How we test

We don’t claim to have reviewed every AI tool on earth. A badge is a record of a specific check, with a date.

The catalog check behind this version of the site ran on October 15, 2026. Counts below are how that pass came out. They are not user totals, ratings, or a quality ranking.

Tools checked
45
Smoke-tested
6
Docs checked
37
Not fully checked
2

Free tier on that pass: 28 yes, 2 no, 15 unconfirmed. Unconfirmed stays unconfirmed.

What the labels mean

Smoke-tested

We ran a real local install or an equivalent hello-path in this environment, such as pip install plus an import or a CLI help command. No paid plan, no account signup, and no model or API call unless a free local path exists without keys. The tool page names the package version and the command.

Docs checked

The official site and/or docs and pricing pages were fetched well enough to record the URL status, pricing or free-tier notes, and a public release date when one existed. There was no successful local hello-world. This label is visible on purpose so it cannot be mistaken for a hands-on test.

Not fully checked

We could not finish docs and pricing verification. A homepage blocked with HTTP 403 is the usual reason. Those tools stay in the catalog with that label. They are not placed on a use-case short list.

Checklist

  1. Site loads. HTTP GET the canonical homepage and follow redirects. Record the status. A 403 from bot protection is not proof the site is dead.
  2. Pricing and free tier. Prefer the official pricing page. Record whether a free tier, free plan, or community self-host edition is explicit. If it is unclear, mark it unconfirmed. Never invent a price.
  3. Docs. Confirm a docs URL responds, or use the GitHub README for an open-source project.
  4. Latest release. For GitHub-backed open source, use the GitHub Releases API. For closed products, say when no single public release feed was used.
  5. Hands-on, where a local install is feasible. Install in an isolated environment and import or run help. Do not create accounts or pay. Do not claim a model run without keys.

What we did not do

  • We did not sign up for any product or enter payment details.
  • Hands-on Python installs did not invoke models. No API keys were used.
  • npm install n8n was attempted and failed in this environment. That failure is on the n8n page, and the badge stays Docs checked.
  • HTTP 403 to our crawler is an environment limit, not proof a browser fails.
  • We do not sell the smoke-tested badge.
  • Paid “Featured” placement, when it exists, will be labeled Featured (paid) and is separate from testing. None of the current listings are paid placements.
  • There are no star ratings, testimonials, or user counts on this site.

New this week

“New release” means a public release timestamp on or after October 1, 2026, relative to the October 15, 2026 check. “New this week” on the homepage is the tighter window starting October 8, 2026. The directory itself was seeded on October 15, 2026, so we do not mark every listing New just because the site is new.

See the current list. The same rules are summarized for agents in llms.txt.