Skip to content
New: Self-serve pilots are open · Upload documents, get a design review
FedRAMP Class A · SOC 2 Type II

Read what actually governs.

Specset merges the base spec, every addendum, and the special provisions into one conformed text per section, resolves which document rules where sets overlap, and answers questions with the paragraph and page behind every claim.

One governing text per section. Every addendum merged in place, every change traceable.
01 · Conformed

One governing text per section.

Base spec, addenda, and special provisions merged in place, with the diff kept for every revision.

02 · Paragraph-cited

Answers land on the page.

Every claim cites its paragraph, and every citation opens the section where the language lives.

03 · Exhaustive

Coverage by construction.

Scans visit every section with its own worker. No sampling, no “top results”, no missed division.

Trusted by the teams delivering public work
SkanskaHNTBManson ConstructionTriton Marine ConstructionGuardian ConstructionPacific Tech ConstructionTurnagain Marine ConstructionGhirardelli Associates
Half the risk was in Addendum 3. Specset had already merged it into the sections before anyone on our team opened the base spec.
[Name] · [Title] · [Firm]
31K+Specification sections conformed and indexed
2.6M+Pages read into project records on Specset
01 · How the set is built

Structured, conformed, governed.

The dangerous requirement is rarely in the base spec. It is in the addendum that modified it, in the set nobody re-read. Specset does the merge before anyone reads anything.

02 · Ask it, scan it

Every section answers for itself.

Ask a question and get the governing paragraph back. Or run a scan that walks all of Division 01 through 33 and returns the requirement register you actually needed: warranties, testing, hold points, cited row by row.

Common questions

What teams ask about spec intelligence.

01What spec formats does it handle?

MasterFormat and CSI project manuals, plus agency formats like Caltrans, WSDOT, FDOT, UFGS, and BNSF, including special provisions amended into standard sections. Each document gets its own parsing pass, so owner-specific quirks are handled per set rather than forced into one mold.

02How does conforming actually work?

The system walks the base section and its amendments oldest to newest and applies the directive language as written: replace this value, add this paragraph, delete that subsection. The result is one governing text per section, node by node, with the revision history and an inline diff kept for every change.

03What happens when a new addendum lands?

Affected sections re-conform automatically into a new revision, and you can diff it against the previous one to see exactly what moved. Everything downstream reads the updated text immediately: search, reviews, registers, and answers in the thread.

04Which document governs when sets overlap?

Sets are ordered by issue date and files within a set by precedence, so when the same section number appears twice, the later governing text wins. Building-area overlays supersede the master spec where they apply, and the agents are taught not to flag that as a conflict.

05Can we pull every warranty, test, or QC requirement out of the specs?

Yes. A scan visits every section with its own worker, no sampling, and returns a table where every row cites its paragraph: warranty durations, testing and inspection requirements, QC hold points, whatever the question is. The table exports to a spreadsheet.

06How do citations work?

Every paragraph carries a stable node anchor tied to its page in the source PDF. When an answer or a finding cites 1.5.A, clicking it opens the conformed section at that paragraph, with the trail back to the file the language came from.

Start with your current set

Put the whole set in one place.

Send a project manual and its addenda. We'll return the conformed sections with the changes visible, ready to search, scan, and cite.