A portable, attest-able identity for autonomous agents independent of their host app.
Repository: https://github.com/AIFrontiersLab/innova-agent-identity-passport-c37d24
Agents are currently just API keys inside someone else's account.
Why now: Multiple stacks are inventing incompatible agent IDs at once.
What recently changed: A2A-style protocols and tool-using agents in production pilots.
Why it was hard before: No persistent non-human actors with cross-system actions.
| Future thesis | Without identity, nothing else in the agent stack (auth, pay, insure) composes. |
| First customer | An agent platform that needs audit-grade actor IDs |
| Market | Identity infrastructure |
| Convergence | Agent Identity, Authorization |
| Moat | Issuer network + revocation + sponsor graph. |
Issue a DID for an agent, bind it to a sponsor, sign actions, verify elsewhere.
- Accept an agent (or human-on-behalf-of-agent) intent
- Enforce scope / policy / evidence
- Execute the smallest useful autonomous step
- Leave an audit trail a human can kill-switch
Issuer, resolver, revocation list, sponsor binding, action signatures.
python3 -m http.server 4173 --directory "/Users/aiserver/Workspace_New/VectraFutureLab/outputs/futurelab/2026-08-21/run_f54ec8b8365f/prototype"It is the substrate other primitives require.
Too infrastructural; no buyer until a crisis.
- Standards freeze around a big-tech format
- Governance of issuers
- Service accounts
- API keys
- workload identity
- Cross-vendor agent actions will matter
This project is intentionally scoped to AI and Agentic AI only.
ai agentic-ai autonomous-agents vectra-futurelab prototype agent-identity authorization