Live
For agencies, consultants and freelancers
Writing a proposal is mostly re-typing what you already know, then
arguing with a blank page about the wording. CM Proposal AI takes a
short questionnaire about the project and returns a structured
proposal — scope, timeline, costed investment table — that you edit
section by section and export as a branded PDF.
- A four-step questionnaire instead of an empty document
- Every section editable, because a draft is a starting point
- An investment table whose line items have to add up
- Branded PDF export using your logo, colour and payment terms
Live
For agencies, and anyone with a business website
Paste an address and get a report you can hand to a client or
straight to a developer. Every finding says what it costs and
exactly how to fix it — the tag to add, the header to send, the
number to aim for. Add your own name to it and print it.
- Titles, descriptions and link previews, with the fix for each
- Structured data, and whether it actually parses
- Alt text, HTTPS, mixed content and security headers
- Prints as a branded, client-ready document
- Optional weekly re-check, emailing only when something changes
Live
For any business with a website
Structured data is how a search engine — and increasingly an AI
assistant — knows what your business actually is, rather than
inferring it from your page text. This builds valid JSON-LD from a
short form, in your browser, and nothing you type is sent anywhere.
- Business type, address, opening hours and services
- FAQ markup, with a warning to keep it matching the page
- Validated as you type, with errors separated from suggestions
- No ratings or review counts — inventing those risks a penalty
Live
For oil & gas operators, importers and exporters
Article 30 required the EU Methane Transparency Database to be
running by 5 February 2026; it was not, and launch has been reported
for September. From 2027 new import contracts must show measurement
and verification equivalent to EU producer standards. Most operators
can describe their programme; far fewer can evidence it against each
obligation with a date attached. Eight questions gives you the list.
- EU Methane Regulation, OGMP 2.0 and US GHGRP obligations
- Every finding cites its source, so it can be checked
- Ordered by which deadline falls due first
- Runs in your browser — your compliance posture stays with you
Live
For shipowners, charterers and bunker buyers
LNG in a medium-speed Otto engine clears the 2025 GHG intensity
limit by 0.13 gCO2eq/MJ and misses 2030 by 3.51 — because Annex I
counts fuel that leaves the engine unburned as methane, at 25 times
the weight of CO2. Put your own fuel mix in and get the penalty from
the regulation's own formula rather than from a model.
- GHG intensity, compliance balance and the Annex IV penalty in euros
- EU ETS allowances alongside it, on the same voyages
- Every factor cited to the Official Journal, including the chain to RED II
- No invented allowance price and no invented biofuel factors
Live
For oil & gas operators
135 prompts mapped onto the oil and gas value chain — drilling,
production, subsurface, reliability, HSE, regulatory, emissions,
supply chain and commercial. Each covers a job a competent person
does today, and each is a prompt you copy into your own model.
This is a library to work from, not software that runs.
- Nine functions across 41 job trees, around one operator knowledge base
- Every prompt states what it needs, what it returns and its guardrails
- Guardrails written for process safety — reporting only, never barrier decisions
- A build ladder, so it is a sequence rather than a wish list
Live
For anyone retyping numbers off PDFs
Drop in an invoice, receipt, contract or CV and get the fields back
as data. The part that matters is what happens when a field is not
there: it comes back empty with a reason rather than filled in with
something plausible. A wrong total that looks confident is worse
than no tool at all, because nobody re-checks a tidy number.
- Every value quotes the exact text it was read from
- Flags what it could not verify — cropped scans, totals that do not add up
- Your own field names, not a fixed template
- The file is read and discarded — never stored, never logged
Live
For anyone answering the same questions all day
A bot that invents your returns window sounds exactly like one that
read the policy — same tone, same certainty, nothing for a customer
to detect. This one can only say what your policy says, quotes the
passage it used, and the server checks that quotation is really in
your text before you see it.
- Answers quote the policy, or there is no answer
- Fabricated quotations are caught server-side and withheld
- Questions it cannot answer become a drafted handover, not a dead end
- The gaps it finds are your documentation backlog, written by customers
Live
For deciding what is worth automating at all
Describe something you do by hand every week. You get the steps as
they actually happen, each marked automate, assist or keep human,
with the reason and with what breaks when it breaks. The useful
answer is usually which steps not to automate.
- Hours saved computed in code from your figures, not by the model
- No frequency and duration means no estimate at all
- Every step carries a failure mode, not just a happy path
- Automated steps credited with 90% of their time, never 100%
Live
For anyone who needs a site that does not look like a template
Describe a brand and get a complete scroll-film site — one
continuous journey with its own concept, palette, type pairing and
written copy, as a single file you download and host anywhere. Two
briefs produce two worlds that look nothing alike, because there is
no theme underneath being re-skinned.
- The model art-directs and writes; a tested renderer builds the page
- Contrast and motion variety checked before you ever see it
- Reduced motion respected, skip link past the film, real footer
- One self-contained file — no account, no lock-in
Live
For anyone who owes the internet a month of posts
A brand kit, a vector logo and a dated calendar written separately
for each platform — the same idea said differently, not one caption
pasted five times. Approve what you want, bin what you do not,
export the rest.
- Captions measured in code against the real limits, hashtags included
- Dates computed, not left to a model that is right almost every time
- Nothing scheduled until a person approves it
- No reach estimates, no engagement predictions, no invented numbers