The library,
in your pipeline.
Your warehouse already ingests the ERP, the DAM, and what Revit exports. What it doesn't have is materials data. Norma API feeds it a materials master — entity-resolved, cited, re-verified — the same firm library that runs by email and inside Claude, callable from your own stack and your own agents. Not generally available yet.
You have a warehouse. You don't have materials data.
Snowflake or Databricks is running. ERP facts land daily — POs, invoices, budgets. The DAM holds every asset, and something is already harvesting Revit. Then the joins fail, because none of it agrees on what a product is: no clean product, vendor, or material dimension anywhere in the stack. That's the gap Norma fills.
The API doesn't replace the firm library — it exposes it. Your warehouse subscribes to a certified source: golden records with lineage on every fact, kept fresh by the same agents that work your projects by email.
If you don't run a data warehouse — or your own tooling — you don't need this page. Email is the front door, and it stays that way.
A source that arrives gold, not bronze.
Everything else lands in the lake raw, and your data team pays to clean it on the way to the warehouse. The library skips the lake — it arrives entity-resolved, cited, and re-verified: the dimensions your fact tables have been missing. And the master writes itself: designers forward quotes and ask in Claude, the agents parse, resolve, and cite — no backfill project, no data-entry team.
ERP ──────────── POs · invoices ──────▶ ┌───────────────────────────┐
DAM ──────────── assets · cutsheets ──▶ │ YOUR WAREHOUSE │
REVIT ────────── model exports ───────▶ │ Snowflake / Databricks │
│ │
email threads ─────────┐ │ fact_po │
Claude + Norma MCP ────┤ writes │ fact_invoice │
your own agents ───────┘ │ fact_model_usage │
│ │ │
▼ │ dim_product ◀── new │
┌──────────────────────┐ │ dim_vendor ◀── new │
│ NORMA LIBRARY │ API · sync │ dim_material ◀── new │
│ parse · resolve · ├───────────────▶│ │
│ cite · re-verify │ arrives └─────────────┬─────────────┘
│ canonical · gold │ gold │
└──────────────────────┘ ▼
▲ joins that finally work:
│ corrections flow through spend by material · price
└─ the agent, not the copy drift · vendor performance The warehouse copy is a projection, never the canonical. Reads flow out — a sync, an extract, the API. Corrections flow back through the agent, so identity, citations, and receipts stay intact. Edit the copy and you've forked your golden record; route the fix through Norma and every projection heals.
What you'd build.
Four shapes we keep hearing about. Each one reads and writes the same firm library the email surface uses — one source of truth, your output format.
- 01
A materials-master feed into Snowflake or Databricks.
→ dim_product, dim_vendor, dim_material land in the warehouse — entity-resolved and cited — and spend, price drift, and vendor analytics finally have something to join against.
- 02
A Revit add-in that pulls spec sheets and FF&E rows into the BIM model.
→ Pick a chair family in Revit, see the matching rows, drop in the studio-approved variant with cited spec data.
- 03
An internal sourcing dashboard that queries the library alongside live feeds.
→ A specs lead pulls a budget heat-map across active projects without leaving the studio's tooling.
- 04
A spec-book exporter that turns the library into client-ready PDFs.
→ A working schedule becomes a typeset, branded spec book on demand — same source, your template.
Where the API actually is.
It exists internally — the email surface and the Claude connector both run on top of it. We haven't opened it for external studios yet. We're talking to the first few, and your case shapes the surface.
Two keys, two jobs. A Norma API token — scoped per agent, like the address edgewood@agent.norma.llc — authenticates you to Norma: it reads and writes only its agent's slice, nothing leaks across studios, and the same kill-switch rules apply. Your own model token (Claude or whatever your harness runs, bought in bulk) pays for your orchestration. Bring your own.
Pricing is set with you. Norma API rides on the same library underneath email and Claude — so there's nothing extra to store, just a way in. We price it per engagement in early access; tell us your shape and we'll quote it.
It's not GA. The fastest path is a short note about your tool, your stack, and the operation you need. If we're not ready for your case, we'll say so.
Tell us what you're building.
One paragraph on the tool, the stack, and the operation you need. Same library underneath the email, the connector, and your code.