Your archive knows. Now you can ask it.
Years of contracts, offers, invoices, correspondence and drawings sit on your NAS. nasindex makes them searchable for AI. You ask in plain language and get the answer together with the original document.
With Claude, or fully local with your own AI, no cloud at all.
Project 24031-MST holds three offers for BKP 22810, sun shading on the south façade. I read all three and put the amounts side by side.
The answer is in the archive. Somewhere.
Looking for an old offer or contract means clicking through project folders and subfolders. Knowing how many offers came in per supplier or trade means counting by hand. The knowledge is there, but it costs time every single time.
«What did that cost back then?»
The figure is in a PDF. Nobody quite remembers which one.
«Where else does this company appear?»
Searching across all projects means opening every folder.
«Where are we missing a contract?»
Gaps only show when you lay everything side by side.
The AI asks, nasindex finds.
nasindex opens up your archive once and keeps it current from then on. The AI finds the matching document, reads it and answers with the source.
| You ask | nasindex delivers |
|---|---|
| What was offered for the sun shading on the south façade? | The offers for that trade, read in the original |
| How many offers do we have per trade? | A count across the entire archive |
| Which trades have invoices but no contract? | The gaps, project by project |
| Where else does company X appear? | Every occurrence across all projects |
| All contracts over CHF 50,000 | A list with contractor, amount and date |
Built for figures that have to be right.
Your archive stays untouched
nasindex only reads. It moves, changes and deletes nothing. Access runs through a user that can only read.
Every figure from the original
Amounts and dates always come from the original document, never from a summary. Every answer names its source.
Your data stays with you
Whatever nasindex builds stays on your NAS. Access is only possible from your private network and with a key of your own. With Claude, only what Claude reads for a question goes to Claude. With a local AI, nothing leaves your network.
Your folders stay as they are
nasindex works with the folder structure you already have and uses it for search. Nothing needs to be rearranged.
Fully private. No frontier models.
nasindex does not need a cloud AI like Claude or GPT. You can open up your archive entirely inside your network and query it with your own, locally run AI. Then no document, no question and no answer leaves your building.
Without internet
nasindex reads even scanned documents right where they are. Nothing goes out in the process.
Any AI that speaks MCP
Any MCP-capable client with a local model can search, read and retrieve originals, just like Claude.
Contracts with your model
If you wish, evaluating contracts also runs on your own model.
For Claude everything is ready to go. A local AI must be able to call tools, and how well it answers depends on the model. We work out the right combination for your hardware with you.
Amounts from contracts, as a list.
For questions about thresholds and totals, nasindex reads the key details from every contract: contractor, subject, date, amounts, signatories and contact.
This only runs on your command and only with your approval. You choose which language model reads the contracts, if you wish your own model inside your network.
Reference value from a test run with Anthropic. With Anthropic you see the estimated cost for your archive before every run.
Three steps, then it runs.
Install
You receive a package. A wizard asks a few questions, checks access to the archive and starts nasindex. On the NAS itself or on a computer that runs around the clock.
Let it open up your archive
The first run happens on its own, a few hours for a large archive. After that, whatever is new or changed is added continuously.
Connect your AI
Connect Claude Desktop or Claude Code once per workstation, or your local AI. Then start asking.
For firms with an archive that has grown.
For anyone whose contracts, offers, invoices, correspondence and drawings have grown on a NAS over the years. nasindex gets the most out of construction project archives: architecture and site management firms, general contractors and owner's representatives whose filing is organised by project, BKP and award.
You need
- A NAS with Docker, for example Synology with Container Manager, or a Linux computer that runs around the clock
- Tailscale on that device and on the workstations
- Claude Desktop or Claude Code, or an MCP-capable client with your local model
To be honest
- CAD drawings and images are not read for their content. They can be found by folder and file name.
- New files usually appear within an hour.
Before you ask
Does nasindex change anything in our archive?
No. nasindex only reads, through a user without write rights.
What data does Claude see?
Only what Claude searches for and reads for a question: search hits and the documents it opens. Everything else stays on your NAS.
Does it work without any cloud AI?
Yes. nasindex opens up your archive entirely inside your network. You can query it with any MCP-capable client and a locally run model. Evaluating contracts can also run on your own model. Then nothing leaves your network.
Do we need new hardware?
Usually not. A NAS with Docker is enough for permanent operation. For the first run over a very large archive a more powerful computer can step in, after which the NAS takes over.
Do we have to restructure our folders?
No. nasindex works with the structure you have. The more clearly it is organised, the more it gets out of it.
Is it only for construction projects?
No. Search works for any document archive, including scanned documents. In construction project archives nasindex also recognises trades and award stages and can evaluate contracts.
What happens if the NAS is briefly unavailable?
Search keeps working, and nasindex loses nothing.
What does nasindex cost?
We discuss price and licence with you in a call.
Ask your archive its first question.
In a call we look at how your archive is structured and what nasindex makes of it.
Book a call