Aegis
Blockchain forensics & intelligence platform
A self-hosted investigation platform for law-enforcement and government analysts to trace stolen or illicit funds across chains, flag scam entities, and produce AI-assisted investigation briefings — designed to run air-gapped, because the addresses an investigator looks up reveal who they're pursuing.
- Pathfinder trace engine: seed an address and it walks the money-flow graph N hops outward automatically, stored as an openCypher graph on Postgres (Apache AGE) behind a swappable GraphStore interface.
- Watchtower scam flagging: flags are published as events over NATS JetStream; flagged entities are marked across every open trace graph and pushed live to investigators via Socket.IO.
- Scribe AI analyst (Claude): grounded case Q&A plus a pattern assessment against 16 fraud typologies — peel chains, mixers, bridge hopping, pig-butchering, drainers, address poisoning and more — with structured, schema-validated output and prompt-injection hardening.
- WebGL investigation canvas (Sigma.js + Graphology) that lays hops out as bands and collapses repeated transfers into weighted edges, keeping large traces readable.
- Security by design: per-case AES-256-GCM envelope encryption, a SHA-256 hash-chained tamper-evident audit log, JWT + role/permission guards and service tokens.
- Multi-chain adapter layer on viem with CAIP-2 chain ids, metered-RPC → public-RPC fallback and block-range chunked log scanning (Ethereum, Polygon, Base, BSC).
An Nx monorepo of NestJS services that talk over NATS events, with shared libraries for auth, crypto, chain adapters and the graph store.