ArkhamMirror: SHATTERED is a modular, air-gappable, local-first and AI-assisted platform. It can run entirely on your machine - no cloud, no telemetry, no ads, no subscriptions, no fees, no limits.
It’s designed for journalists, researchers, legal advocates, OSINT enthusiasts or anyone else who needs to find truth in documents but can’t (or won’t) expose them to Big Tech and can't afford to use Palantir.
Key features:
- 25 modular plug-and-play "shards" (CIA's Analysis of Competing Hypotheses tool, MOM/POP/MOSES/EVE techniques, entity extraction, timeline, graph visualization, contradiction detection, etc.) attach to the central "frame" - PostgreSQL-only infrastructure (pgvector for embeddings, SKIP LOCKED for job queues) - Runs on 4GB RAM using cloud embeddings, add ~1.2GB for local models - Full audit trails for evidence provenance - Dockerized for easy isolation and/or VPS deployment - Use pure local inference for sensitive data or use any OpenAI-compatible cloud endpoint for unclassified materials, your choice. (I personally use Anthropic inference through OpenRouter's BYOK option for cloud services and LM Studio's Qwen3-vl:8b for local inference)
Use cases we've focused on: investigative journalism, OSINT, legal self-advocacy, healthcare insurance appeals, civic oversight.
The "Healthcare Self-Advocacy" section exists because insurance companies are deploying AI to deny claims faster. This tool helps people fight back. There are 25 individual shards now, but more shards are in development.
MIT licensed. No VC funding. No growth metrics. Just trying to give powerful tools to people who need them. Intelligence for anyone.
I'm more than happy to answer any questions and welcome suggestions, criticisms, bug reports, you name it.