◆ Autonomous Worker
Ingests SSI updates, validates them against the counterparty, retires stale accounts, and keeps the directory clean. A wrong SSI propagates downstream as a fail and a break, so every change is counterparty-confirmed before publish and publishes are versioned and reversible.
Memory
Working The incoming SSI change vs. the current directory entry.
Episodic Recent fails/breaks attributable to this counterparty's data.
Semantic SSI structure by market, account-format rules, counterparty hierarchy.
Store Versioned static-data directory
Orchestration
MCP
Harness · Managed Agents … event-driven + scheduled runs; deterministic validation.
Tools
{ } SSI / static-data directory API ⌘ Counterparty confirmation channel MCP ›_ Format / account validator Code exec
Evals & guardrails
- Every SSI change validated + counterparty-confirmed before publish; bad loads quarantined.
- Versioned, reversible publishes with diff review on bulk changes.
- Stale-account detection cross-referenced against the fails feed.
Frontier edge
- ▲Proactive: predicts which directory entries will trigger fails (drifting from the fails feed) and pre-emptively re-validates them before they break anything.
- ▲Agent-mesh negotiation: confirms ambiguous SSI changes directly with the counterparty's static-data agent over A2A instead of an email round-trip.
- ▲Formal action-gating: every publish is a cryptographically signed, replayable directory mutation that provably stays within the validation policy.
In numbers
99.97%
SSI accuracy
−41%
Data-driven fails
Handoffs
Hands to → Settlement Instruction Agent