For agents / data
Data as of 2026-10-18 (Scoring rules v0.1). Informational only; not a guarantee, endorsement, or certification.
The directory is available as plain files that programs and AI agents can read directly. The files are generated from the same approved entries as the pages and contain the same scores and grades.
How the counting and scores work: see How this directory works.
Files
- agents.json — all entries with factor scores, grades, source counts and cited sources.
- agents.csv — one row per entry, without the source lists.
- schema.json — JSON Schema describing agents.json.
- llms.txt — short summary and links for language models.
- llms-full.txt — all entries in one plain-text file, generated from the same data.
Fields
| Field | Meaning |
|---|---|
| id | Stable identifier (the entry slug, e.g. elicit). Never reused; withdrawn entries stay with status withdrawn. |
| name, url, homepage | Display name, the entry page on this site, and the agent’s own homepage. |
| category, type, scope | Category, commercial/open-source type, and what the entry covers. |
| status | Scored or Insufficient evidence (fewer than three independent performance sources; no overall score). |
| scores | The four factors, 0–100 each, shown separately (completion rate 30%, disputes 20%, adherence to limits 25%, evidence quality 25%); empty/null when evidence is insufficient. |
| overall_score | Weighted overall score, whole number (exact .5 rounds up); empty/null when evidence is insufficient. |
| evidence_grade | A, B, B- or Insufficient evidence. Describes the evidence, not the agent. |
| independent_source_count / counted_performance_sources | Independent performance and other sources as stated in the entry, and the number of counted performance sources. |
| last_reviewed | Date the entry was last reviewed (America/New_York). Values are as of this date. |
| sources (JSON only) | Each cited source: url, kind, date accessed, whether counted, whether peer-reviewed. |
| disclaimer, license | Repeated on every CSV row so copied rows keep them. |
Usage note
If you use or quote these scores: they are informational, rule-based assessments of public evidence as of each entry’s last_reviewed date (Scoring rules v0.1). They are not guarantees, endorsements, certifications, or rankings of overall quality. Quote the factor scores and evidence grade together, include the as-of date, and link the entry page. Do not describe an entry as ‘verified’, ‘certified’, ‘approved’ or ‘trusted’ by the Clearinghouse.
License and citation
The data and the site text are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You may share and adapt them, including commercially, if you give credit. Cited third-party sources remain under their own terms. The code (scripts/, the stylesheet and vercel.json) is licensed under the Apache License 2.0; CC BY 4.0 covers the data, the decision records and the site text, including generated files. The project name, logo and favicon are not licensed. See LICENSE-DATA.md and LICENSE.
Suggested citation: Agent Reputation & Trust Clearinghouse, <entry URL>, data as of YYYY-MM-DD, CC BY 4.0.
Ready-to-copy citations (plain sentence, APA-style, BibTeX, agent template) and how to quote a score: How to cite and quote this data.
Updates
The files are regenerated whenever an approved decision changes an entry; each change is listed in the changelog. The same changes are published as an update feed in JSON Feed 1.1 and Atom formats; each item has a stable id, date, type, agent id, before and after score, grade and status (null where not recorded), decision packet ids, the entry URL, an as-of date and the informational disclaimer. There is no fixed schedule, and copies held elsewhere may be out of date. This page is the canonical source.
Corrections
Corrections follow the same process as for the pages: see Methodology: Corrections or email trustkeeperace@gmail.com.
Disclaimer
Informational only; not a guarantee, endorsement, or certification. Scores reflect public sources and the stated rules as of each entry's last_reviewed date. The Clearinghouse is not responsible for any agent’s outcomes or for how others use this data.