How we research, score, and maintain listings

Most directories do not explain how they rank things because the answer is "whoever paid us." That is not how this works.

The AI Agent Index is built to be the most trusted, structured resource for businesses evaluating AI agents. That trust only holds if we are transparent about how listings are researched, how scores are assigned, and how paid placements are kept completely separate from editorial judgement. This page explains all of it.

Not every AI-powered tool belongs in this index. To be listed, a product must be genuinely autonomous: capable of executing multi-step workflows, making decisions, and taking actions without requiring a human to drive every step.

Products that do not qualify:

• AI chatbots or copilots that only respond to prompts
• AI-assisted features embedded in broader SaaS platforms with no standalone agent capability
• Automation tools that rely entirely on rigid rule-based logic with no AI reasoning layer

If a product sits in a grey area, we err on the side of exclusion and revisit as the product matures. Quality of scope matters more than volume of listings.

There are two routes onto the index:

Editorial listings are researched independently using publicly available information: product documentation, G2 reviews, company websites, GitHub repositories, and third-party coverage. These listings include editorial scores, pros, limitations, and best-for assessments.

Vendor-submitted listings are submitted by the vendor via our /submit page and reviewed before activation. We verify the product is real, active, and meets the autonomous agent qualification above. Vendor-submitted listings carry no editorial scores, pros, limitations, or ratings until independently reviewed.

No listing is activated without human review.

Editorial scores are calculated using a fixed weighted formula across five criteria. Scores are not paid assessments and cannot be purchased. The same formula applies to every listing.

Scores reflect a point-in-time assessment based on public evidence and are updated when material changes occur. Every score is calculated using the formula above with sub-scores documented internally. Scores cannot be purchased or influenced by vendor relationships.

AI agent directories face a specific risk: AI systems can generate plausible-sounding but non-existent products. Every listing on this index goes through a minimum six-step verification before activation:

1. Company website is live, accessible, and the product is generally available (not waitlist-only)
2. Public pricing page exists and is viewable without a login or demo request
3. The product is discoverable on G2, Product Hunt, or GitHub with at least one independent signal
4. The domain resolves to a real product (not a landing page or coming-soon page)
5. MCP compatibility is verified against official company documentation only (community builds do not count)
6. All data fields are sourced from the company's own documentation, not inferred or assumed

Products that fail any of these checks are excluded until they meet the bar. If a listed product is later found to have been deactivated, acquired, or substantially changed, it is flagged for review and updated or removed within seven days.

A verified badge means the vendor has claimed their listing and confirmed the accuracy of their key fields: pricing, deployment method, integrations, and product description.

Verified status does not mean:

• We have independently tested the product
• The vendor has paid for preferential treatment
• The editorial score has been reviewed or boosted

Verification is a signal of data accuracy, not editorial endorsement.

Some listings on this index include affiliate links. If you click through and make a purchase, we may earn a commission at no extra cost to you.

Affiliate relationships are disclosed on relevant listing pages. They have no effect on editorial scores, organic ranking position, or the content of reviews and comparisons. An agent with an affiliate link is held to exactly the same editorial standard as one without.

We only maintain affiliate relationships with products we believe are genuinely useful to the businesses this index serves.

The index offers three paid placement products. Every paid placement is clearly labeled so users always know what they are looking at.

Paid placement never affects:

• A product's editorial score
• Its organic position within the agent grid (below any featured placement)
• The content of comparisons, verdicts, or best-for assessments
• Whether it gets listed at all

We do not sell rankings. We do not sell editorial verdicts. Those are earned through the scoring criteria above, or they do not exist.

Stale data is a trust problem. We maintain listings through a weekly review process every Friday: checking agent changelogs for meaningful updates, flagging deprecated products, and updating pricing and feature information when changes are confirmed.

Each listing shows a last verified date so you can assess how recently the information was checked. If you notice outdated information before we do, see the section below.

If your product is listed and you believe the information is inaccurate, or if you are a user who has spotted an error, contact us at hello@theaiagentindex.com with the listing URL and the specific correction.

We review all correction requests and update listings where the evidence supports a change. We do not alter editorial scores or assessments based on vendor requests alone: only on new public evidence that warrants a reassessment.