Roadmap
Shipped (v0.1)
- 5-pillar trust scoring algorithm (Identity, Safety, Reliability, Transactions, Age)
- Agent registry with full lifecycle (register, claim, verify, kill-switch)
- Red-team safety probes (automated prompt injection testing)
- Escrow create/settle with trust score feedback loop
- Health metrics ingestion + probe worker
- Audit logging on all mutations
- REST API with API key auth
@agntor/sdkon npm@agntor/mcpwith 13 tools (Claude, Cursor, Cline)@agntor/trust-proxyExpress middleware- EIP-8004 agent card endpoint
- AI Judge (Gemini 1.5 Pro) for escrow disputes
In Progress
- Health probe worker deployment to GCP Cloud Run Jobs
@agntor/trust-proxynpm publish- Docs overhaul (trust algorithm spec, API reference, quickstart)
Next (Q1 2026)
- On-chain escrow integration — connect AgntorEscrow.sol settlement to trust score updates
- Multi-framework SDK examples — LangChain, CrewAI, Vercel AI SDK, AutoGen integration guides
- EIP-8004 formal submission — standardize the agent-card JSON format as an Ethereum EIP
- Trust score decay on all pillars — not just safety; reliability and transactions should decay with inactivity
- Webhook notifications — alert on score changes, escrow settlements, probe failures
Later (Q2-Q3 2026)
- Multi-registry federation — let other registries compute trust scores using the same algorithm
- Compliance certification exports — SOC2-compatible audit reports from audit_logs
- Agent-to-agent trust negotiation — agents exchange trust proofs before transacting (no central API call)
- On-chain identity anchoring — link agent IDs to ENS / wallet signatures