What's new

Changes, dated.

Most recent first. Only things that shipped; what's planned is on the plans.

2026-09-23

Connect your own AI, in the Library. For organisations that already run an AI they trust: an administrator makes a key, and that AI can search the sheet index and the text on the sheets through Sheetsa's MCP server. It receives records only, never a drawing or a link to one, and only your organisation's; every request is listed; revoking every key closes the door. Offered beside "Ask the archive" so an organisation chooses by its own security needs. What is shared.
Ask the archive, in the Library. A question in plain words ("which sheets show the pumps serving the penthouse air handlers?", "where is AHU-1?") is answered from your organisation's sheet index and the text printed on each sheet, with every line citing the sheet it came from, its read state and the words that matched; a citation opens the sheet in its set. Sheets' text is read in your browser when a set is added, or afterwards with "Read the text for questions". Answered inside your browser and your organisation's own database: no AI company receives your drawings, and no language model is involved today; when one is added it will run where your data already is and see only the matching rows. What is stored.
Package and send, in the Library. Tick sheets in a set or in search results, or take a whole set, and Archive packages them with their index (JSON and CSV) and a manifest with a hash per file into a link that opens without an account, for 7, 30 or 90 days or until you close it. Send the link from your own mail program with the message filled in. You choose whether the link allows printing and downloading as well as viewing on Sheetsa; recipients say who they are before the sheets show, and never edit anything. A transmittal PDF listing what was sent, to whom and until when goes into the package and stays with your Sent list, where the recipients' activity (opened, viewed, printed, downloaded) is recorded. Works whether your documents live in Sheetsa or in your SharePoint library. What is recorded.
Your SharePoint library as the repository, in Archive Pro. The console connects a library self-serve; new sets can be written into it with the index on each document's columns, sets already there are imported in place, and every sheet opens from SharePoint. Any repository that can be set up self-serve belongs in Archive Pro; the rest is custom work.
Usage reports from the browser tools. Each run of Split, Specs and the Archive tools now sends a short report: which tool, how long, how many pages, how it went, and counts of what the engine found. Never the drawing, its name or anything read from it. Kept for two years, then deleted. The privacy policy lists every field and has the switch to turn it off; the security page is updated to match.
Share your corrections, if you want to. A box on Split, off by default, sends each correction you make together with the titleblock crop it came from, so the engine can be fixed on exactly the reads it got wrong. What is sent.

2026-09-22

Findable by search engines: every public page now names its canonical address and carries a description and a link preview; a sitemap and robots.txt are published, and the sign-in Library is kept out of search results.
Sheetsa Archive opens for early access. Sign in with your email; the scan restoration, multi-pass OCR and PDF/A tools are free while Archive is in early access. The hosted repository, the current set per building, schedules read into tables and consultant handoff packages are in development, in that order.
The site is rebuilt around Archive and Archive Pro, with Split as the free way to try the reading engine. Design previews of the Archive screens are on the front page, labelled as previews.
Custom solutions for owners, agencies, portfolio and facilities managers get their own page: how an assessment works, what exists today, what gets built, and what stays with your platform.
Terms of use updated for Archive accounts and stored files; privacy policy and security and data handling pages published.
Scanned sheets indexed by OCR in the hosted Library: sheet number and title read off the scan itself, checked against a second reading. Tested on sample scans at scanner resolution; real archive scans are next.
Big sets in the browser, measured: Split now says how large a set a browser tab can take and fails cleanly past it, instead of running out of memory silently.
The package and command line are renamed from pdf-doctor to sheetsa; sheetsa.com goes live.

2026-09-21

Terms of use published for the free tools.
Split's review screen rebuilt to the brand design: a table of every sheet with its read state, a panel showing the titleblock the number was read from, and the engine's reason in a sentence.

2026-09-20

Sheetsa Specs: spec books indexed by CSI MasterFormat section and split into one PDF per section with a front-matter file.
The read states (Observed, Confirmed, Inferred, Verified) shown per field, with the second-opinion OCR read on weak titleblock reads.

Earlier

Sheetsa Split: a born-digital drawing set split into one PDF per sheet, named from each titleblock, with a sheet index you can correct.
Sheetsa Archive tools: restore a faded scan (levels, sharpening, black and white), read it with three-pass OCR that reports agreement, and check a PDF against PDF/A.
Everything runs in the browser with no upload, and offline after the first visit.