Every scored question in the pack, in one pass, inside every word limit.
From your own contracts and results, with a link to the document behind each number and a plain warning wherever you can't prove something yet.
Any AI gets you 80% of the way. The last 20% is what wins.
Takes about 20 seconds
or send us a live packEvery number below comes out of a document this firm already had.
Every page, including the ones the questions were hiding in.
Each with its word limit and its scoring guidance attached.
Real contracts and real performance figures, dated and sourced.
Named now, so you fix them before you submit rather than after.
Nothing was invented. Nothing was guessed. Nothing went in that you cannot point at a document for.
The more you use it, the better it gets — because the library is yours, and it deepens every time you bid.
Below is a real tender pack — 14 documents, 212 pages. Somewhere inside are 23 scored questions, each with its own word limit and its own scoring guidance buried in a different document.
This is an illustration, not a live run. Send us a pack you are bidding on and we will run yours properly.
They come out of the same twenty seconds. Most firms pay someone else for two of them.
Whether this one is worth your time, how much of it you can already evidence, and what is missing — written to be forwarded, not rewritten. The conversation that normally takes a management meeting to have.
Would normally cost you: half a day and a meetingNot “your evidence is weak”. The two specific documents that would move two specific answers, ranked by how much they are worth. You can have someone chasing them this afternoon.
Would normally cost you: finding out at the debriefHow much of this pack you can back with real evidence, before you commit three weeks to it. Sometimes the most valuable answer is that this is not your bid — and knowing on day one is worth more than a polite letter in week six.
Would normally cost you: three weeks, then a lossEvery pack you run sorts and dates a little more of what your firm has done. After four bids you have something you have wanted for years and never had time to build: one place where your track record actually lives.
Would normally cost you: a job nobody ever startsFour reasons. Everything else on this page is evidence for one of them.
Twenty-three scored questions, each inside its own word limit, all in one go. Two to twenty days of work back in under a minute.
Scorers reward hard facts over fine words. Your contracts, your KPIs, your outcomes — the thing no model can invent and no competitor can copy. We will not claim it wins your bids. We will claim the answers are checkable, which is more than most can say.
Where it cannot prove a claim it says so and names what to gather. You are the one signing the submission.
Every pack you run and every bid you win adds to a record that only your firm can see. The tool your competitor uses doesn't get better at your business. Yours does.
You could paste a question into a chatbot today. Six things stop that being the same thing.
Buyers publish the marking scheme and its descriptors, and the scale varies by authority — some mark out of five, some out of ten. Taypra surfaces the buyer's own scale and descriptors alongside each question. Writing every answer to that scheme is the piece we are finishing now.
It draws only on your own contracts, performance figures and results. A tool with nothing to draw on writes something that sounds right — which is the exact thing a scorer is trained to catch.
Where the evidence is not there, it says so and names what to gather. You are the one signing the submission.
Taken from the pack and applied as a hard ceiling, then checked again after writing — because AI cannot reliably count its own words.
Every pack you run and every bid you win adds to a record only your firm can see. A shared tool gets no better at your business. Yours does.
The whole pack in one go, joined up and consistent. Not a chat window and an evening of copy-and-paste.
The best firms do not lose on capability. They lose on time, and on the bid.
No bid manager, no library, no evenings rebuilding the same evidence from memory. The first framework place stops being out of reach.
The repeatable bulk of every bid drafted in a fraction of the time, so your people spend theirs on the commercials, the standout examples and the win themes.
Not just your best writer's. The library does not leave when someone does, and it does not forget the contract you delivered three years ago.
The engine was never built for one sector. It reads whatever pack it is given.
The Procurement Act requires every UK contracting authority to publish every regulated notice through one central feed. We consume the whole feed, not a slice of it — housing, local authority, education, NHS estates, facilities, highways and civils, rail, defence.
Framework terms are fixed — four years, four plus two, five plus one. Award date plus term tells you when a place competition is coming, and that is buildable from public data alone.
Fit scored against your own profile with the reasoning shown — including where the match is only partial. Honest matching is worth more to you than a flattering number.
Where we are selling first is social-housing repairs and maintenance, and the demonstration above reflects that. It is a choice about go-to-market, not a limit on the platform: around 70% of winnable social-housing work runs through frameworks whose competitions are invisible from outside, and the same structure holds across most of the rest of the public sector. One pack reader, one evidence engine, one scoring reader — whatever sector the pack came from.
Every bid tool has a content library. A library is a folder full of old documents. This is an organised record of what your firm has actually done — every fact dated, tied to the document it came from, and marked with where you're allowed to use it.
A completeness score runs between the two, so you can see exactly where you stand from minute one. The gaps are not a failing — they are the list of documents that would make your next submission stronger, ranked by how often buyers ask for them.
How a firm gets from a shared drive and four years of memory to the screen above.
Previous bids won and lost, contract reports, KPI returns, policies, accreditations, case studies. Whatever format they are in. Most firms find this takes an hour of looking, not a week.
We pull out the facts, date them, tie each one to the document it came from, and strip out the marketing language that can't be backed up. This is our work, not yours.
Around ninety minutes. You confirm the facts, correct what is wrong and mark what has expired. Nobody knows your record better than you do, and nothing goes into a bid that you have not approved.
From that point the library is live and it deepens on its own. Every pack you run and every bid you win adds to it — which is why the fourth bid takes less of your time than the first.
Typical first-run time: under a fortnight from sending files to a drafted pack, and roughly two to three hours of your own time inside it. If your records are thin we will tell you before you commit, not after.
We will run it and show you what comes back.
Send one framework pack you are bidding on this month. We will answer every scored question from your own past contracts, show you which ones you cannot evidence yet, and have it back inside two working days. No charge.
Then, if it is useful: a monthly subscription priced on the size of your firm. Published bands, no annual lock-in, a month's notice. Your library leaves with you.
A paid pilot on one framework, one month, no annual commitment. We measure edit time on your pack against your team's own baseline and give you the number, whichever way it falls.
Before you send anything: The data-protection agreement, a named list of every outside company involved, which AI supplier we use and how long they keep anything, proof your data is kept separate, and what happens to it if you leave. Ask and they come by return — we would rather you read them first.
One scored question from a real pack. 500 words, marked against the buyer's own published scheme.
“We have extensive experience delivering responsive repairs across the social housing sector, with robust processes, a highly trained workforce and a strong commitment to quality and customer satisfaction.”
“We deliver responsive repairs for three housing association clients, achieving strong first-time-fix performance and high tenant satisfaction across our contracts, with robust processes for out-of-hours cover.”
“Under a £4.2m responsive repairs contract we completed 18,400 jobs at 96.1% first-time fix against a 92% contractual target, with tenant satisfaction of 4.7 out of 5 across 2,140 responses. Voids returned at 11.4 days against a 15-day target.”
The reason the output is specific rather than competent. Most tools have the left half.
Four things we will put in writing.
Primary storage in London. UK GDPR. Every outside company that touches your data, named.
Enforced deep in the database rather than in the app, and tested automatically every time we change anything.
Not ours, not a supplier's. The system only ever reads your own library.
Tell us about your firm and we'll prepare a live look at the frameworks in your patch before we speak.