Graph Files is the agent-first document context plane. Every file knows what it is, what it connects to, and who touched it — verifiable offline, designed for AI agent fleets.
Every document gets a content_id (cryptographic hash). Same content = same node, across systems. No duplicates.
Documents are nodes with class, context, and typed edges. Find by what something is, not where it lives.
Built for AI agent runtimes. Agents query, classify, and manage documents programmatically — no human UI needed.
Every agent touch sealed with Ed25519 + Merkle proof. Offline-verifiable. Tamper-evident audit trail.
Attaches to Microsoft 365, Google Workspace, AWS. Adds context — doesn't replace your storage.
Learns from agent interactions (Alma memory) and tracks document lifecycle status (Veritas state machine).
Document enters Graph Files. Content is hashed → content_id assigned. Same content always maps to the same node.
Agent reads metadata, assigns class and properties. Node gets typed edges to related documents, projects, entities.
Every agent touch — read, write, classify, move — sealed with an AWP receipt. Ed25519 signed, Merkle proven.
Alma records what worked. Veritas tracks lifecycle. Next agent that touches this document benefits from prior context.
Every document orbits a graph of context, identity, and proof.
| M-Files | Graph Files | |
|---|---|---|
| Designed for | Human users | AI agent fleets |
| Identity model | Object ID in vault | content_id (hash) + node |
| Audit trail | Internal logs | AWP offline-verifiable proofs |
| Intelligence | Aino (human-facing) | Alma + Veritas |
| API design | Desktop/M365 DNA | API-first, multi-agent |
| Cloud attach | Microsoft only | MS + Google + AWS |
Graph Files is in the architecture phase. ADR drafted, dogfood target selected. Domain registered. First MVP in progress.
🚧 Building — Not yet available