onyx-actions
The independent trust & verification layer for the agentic web — the signed check an AI agent runs BEFORE it pays or transacts. Returns a hard PROCEED / REVIEW / HOLD clearance plus Ed25519-signed FACTS any third party can verify offline (tamper -> rejected). Neutral by design — earns nothing from any tx, so it grades what conflicted incumbents structurally cannot. Facts, not judgments.
Listing completeness · 50/100
Presence of published information only. Not a security, reliability or performance score. Same rules for Datoka and external listings.
- Name and traceable source
- 10/10 · Present
- Meaningful description
- 10/10 · Present
- Connection or installation reference
- 20/20 · Present
- Dedicated documentation or source repository
- 0/15 · Missing or insufficient
- Explicit free access or numeric tariff
- 0/20 · Missing or insufficient
- Declared integration protocol
- 10/10 · Present
- Source updated within 90 days
- 0/10 · Missing or insufficient
- Declared capability names
- 0/5 · Missing or insufficient
Evaluated 2026-10-13 · Method listing-completeness-1. Scoring rules
Connect & use
Connect through the publisher’s supported interface. Any credentials or payments are handled by the provider.
A2A agent card
Publisher agent card ↗Categories and declared capabilities
Topic categories are derived from publisher text for discovery, not independently verified capabilities.
The registry does not provide a verified tools list. See the observations, when available, for any tools/list check; a published tools list does not establish that execution succeeds.
Evidence, with its limits.
This listing is indexed. Any domain claim or endpoint observation appears separately with its date and limited scope. No payment execution, factual accuracy or safety certification is inferred.