dwg. ALEX-01 · library of context

Alexandria

Your company's context layer.

Every doc, repo, note, and skill your company knows — indexed for humans and AI agents. Hybrid search, an agent-maintained wiki, and an MCP-native API in one registry.

elev. 1:1 — egyptian blue on papyrus

fig. I — pharos

H ≈ 100cELEV. PHAROSSCALE 1:500 · INK: EGYPTIAN BLUE

ALEXANDRIA

SHEET Nº 1

09 Jul 2026

drawn by the librarian

§ 02 — capabilities

Built as a control plane, not a notes app.

What is canonical, what is stale, what is private — and how agents search, fetch, and write back.

plan nº II — stacks

01

Source registry

Every doc, repo file, note, transcript, and skill — catalogued with type, tier, tags, and freshness. The registry is the source of truth; indexes are rebuildable.

02

Hybrid search

Vector and keyword retrieval over chunked content, with metadata filters. Every hit cites its source so agents and humans know where knowledge came from.

03

Self-maintaining wiki

An agent-written lattice of Markdown pages — entities, concepts, syntheses — that compounds knowledge instead of re-deriving it on every query.

04

Agent API & MCP

One REST surface and an MCP server. ChatGPT, Claude, Cursor, and custom agents query the same context layer your team browses in the library.

§ 03 — elevation

Architecture as blueprint.

Clients speak HTTPS. The API fans out to a registry, object store, hybrid indexes, and an LLM wiki — each independently rebuildable.

sys-01 · context layer

ALEXANDRIA / SYS-01SCALE 1:1 · REV D · PAPYRUSCONTEXT LAYER ELEVATIONClientsWEB · MCP · RESTAPI ServiceCLOUD RUNRegistrySUPABASEObject StoreGCS RAW + WIKIIndexesVECTOR + FTSLLM WikiAGENT MDFAN-OUT → EMBEDDINGS + WIKI MAINTAINER

§ 04 — admission

Open the stacks.

Browse the demo library, or join the waitlist for a workspace of your own.