Documentation
How to drive the workbench.
A practical guide to every studio — what to type, what to upload, what the code does with it, and where you stay in control. Written for engineers and quantity surveyors, not administrators.
Start here
Getting started
Miksir is one account across four studios — Mix Design, LCA, Geo and BOQ. You sign in once and switch between them from a single shell.
Creating your account
Go to the sign-up page and register with an email and password, or continue with a supported single sign-on provider. No card is required to start — every new account can run a 14-day Pro trial, then settles onto the Free plan automatically.
The hub
After signing in you land on the hub — the home screen listing the four studios as cards. Each card opens its studio and shows what you've used against your plan's allowances. Your report history and projects live inside each studio, tied to your account.
The product switcher
Inside any studio, the product switcher lets you move between Mix, LCA, Geo and BOQ without losing your place. Each studio carries its own accent colour so you always know where you are:
The one rule
AI drafts, you confirm, code computes
Every studio runs on the same principle. It is worth understanding before you use any of them, because it explains why Miksir asks you questions instead of silently producing numbers.
- AI drafts. Vision models read your drawings and data; language models write narrative sections. They propose — an extraction table, a suggested parameter, a paragraph of prose — and nothing more.
- You confirm. Every extracted or inferred value is reviewed by you against the source before it counts. Low-confidence reads become specific questions. Unconfirmed data never reaches a calculation.
- Code computes. Quantities, corrections, compliance checks and totals come from deterministic formula engines. The same input always gives the same answer, and every assumption is stated in the output.
Wrong dimensions or rebar counts produce wrong material orders on real sites. Miksir is built so that no calculated result is ever generated from unconfirmed AI output — the numbers you export are computed, auditable, and yours to sign off.
Studio 01
Mix Design
Conversational concrete mix design. You describe the element, exposure and target strength; Miksir walks the code tables and returns proportions with a full compliance checklist.
The chat workflow
Mix Design is a chat, not a form. Open the studio and describe what you're pouring in plain language — for example, "C30/37 for a marine retaining wall, 50 mm cover, 20 mm aggregate." Miksir asks clarifying questions only where the code requires a decision it can't infer, then returns a proportioned mix.
- Describe the element. Structural element, target strength class, exposure environment, workability and any aggregate constraints.
- Answer the clarifying questions. The AI asks where a code table branches — cement type, maximum w/c ratio for the exposure class, minimum cement content.
- Review the mix. You get cement, water, coarse and fine aggregate quantities, the w/c ratio, and a compliance checklist against the governing code.
- Export. Scale from a lab trial batch to a full truck and export the mix and checklist as a PDF.
Standards it follows
Exposure-class requirements — minimum cover, maximum w/c ratio, minimum cement content — are resolved against the standard you're working to:
The mix and its checklist are a design draft. A licensed engineer must review and approve it, and trial batches should confirm strength before production.
Studio 02
LCA
Life-cycle assessment for any material or product. You build a form describing the product, the engine computes the impacts, and you export an auditable report.
Form → calculate → report
- Build the form. Create an assessment with a title, company and functional unit, then add categories and sub-categories. Each line item carries a quantity, a unit, and optional transport (distance in km to site).
- Calculate. The deterministic engine maps each material to its emission factors and computes impacts stage by stage — A1–A3 raw materials and manufacturing, A4 transport, and a documented use-phase estimate — using the EF 3.1 method.
- Read the results. Headline global warming potential (kg CO₂e) and primary energy demand come back with a data-coverage percentage and an uncertainty band driven by a five-point data-quality (pedigree) score. Missing data is tracked, never silently zeroed.
- Generate the report. Export a PDF your client can audit. AI writes the narrative sections; the figures stay computed.
Supplier data-collection forms
You rarely hold every figure yourself. Miksir lets you share a data-collection form so a supplier fills in their own material data directly. Their submission flows back into your assessment — the sender and the person filling it in can be different people — so you collect primary data instead of guessing at it.
Each input's pedigree drives the uncertainty on your headline numbers, from ±5% for a verified EPD in the same region to ±50% where data is unknown. Better inputs mean tighter bands — the report shows you exactly where the gaps are.
Studio 03
Geo
From field blows to design parameters. Paste SPT or CPT field data and get corrected values, soil classification and strength parameters — every correlation named, every step shown.
What to enter
Enter your sounding as a depth profile. For each test point Miksir uses whatever you provide and applies the appropriate corrections; the more you supply, the fewer assumptions it has to make.
| Test | Expected inputs per point |
|---|---|
| SPT | Depth, field blow count N, and where known: hammer energy ratio, borehole diameter, rod length, sampler liner, water-table depth, and a driller's soil description. |
| CPT | Depth, cone tip resistance and sleeve friction, for behaviour-type classification and cone-method cohesion. |
Grain-size class can't be reliably inferred from SPT-N alone. Add a sample description (e.g. "silty sand", "lean clay") or a CPT sounding and the classification is far more confident.
Corrections applied in code
Raw field N is corrected to N₆₀ and then overburden-normalised to (N₁)₆₀ before any strength parameter is derived:
- Energy ratio — normalised to 60% (assumes 60% if the hammer energy isn't given).
- Borehole diameter — Skempton (1986).
- Rod length — Youd et al. (2001).
- Sampler — liner / no-liner correction.
- Overburden — normalised with the correction factor capped at 2.0.
From the corrected values the engine classifies the soil first, then exposes only the physically meaningful parameters — friction angle φ for coarse-grained soils, undrained cohesion cu for fine-grained — plus unit weight and bearing capacity by Terzaghi, Meyerhof, Hansen or Vesic. Published correlations carry ±10–20% scatter, so every result ships with an uncertainty band, and inputs outside a correlation's validity range raise an extrapolation flag.
These are preliminary, advisory estimates for site characterisation — not a substitute for site-specific laboratory testing on final foundation designs.
Studio 04
BOQ
Drawings in, priced bill of quantities out — with a confirmation step that keeps AI out of the arithmetic. Concrete, rebar and formwork are always computed by the formula engine, never by the AI.
The full workflow
- Create a project. Set the title, location, currency, the governing rebar code (
ACI 318,EN 1992orTS 500) and the lap factor (40–50 × diameter). Wastage percentages for concrete, rebar, masonry and roof default sensibly and can be overridden. - Add elements — manually or by extraction. Type structural elements, masonry walls and roof items directly, or upload a structural drawing PDF and let the AI extract every column, beam, slab and footing. Manual entries are confirmed by default; extracted elements always arrive as pending.
- Review and confirm. Extracted values are shown beside the source sheet, with low-confidence reads highlighted amber. Anything unclear becomes a specific, one-line question grouped by sheet and element — "Sheet 3, C4: how many 16 mm main bars?" You confirm, correct or exclude each element. Answers that look physically implausible trigger a follow-up rather than being accepted silently.
- Set rates. Enter unit rates per item key — concrete grades, rebar diameters, formwork, block types, mortar, roof items.
- Calculate. The deterministic engine runs on confirmed elements only. Pending items are listed separately and never priced.
- Export to Excel. A grouped, priced BOQ with subtotals, an assumptions sheet, and a pending-items sheet.
What the engine calculates
- Concrete — volume per element shape (rectangular / circular columns, beams, slabs, pad and strip footings), with wastage stated.
- Rebar — unit mass
0.00785 × d²kg/m; bar length = straight length + lap (lap factor × diameter); bar count by spacing where given. - Formwork — contact area per element shape.
- Masonry & roof — entered directly by the quantity surveyor: wall volume less openings over 0.5 m², block counts from face area, mortar split by mix ratio; roof items pass straight through as priced line items.
Extraction sheet allowances
One extraction sheet is one page of a drawing PDF processed by the AI takeoff. Manual entry is unlimited on every paid plan and never uses your allowance. Extraction is metered per month:
| Plan | Extraction sheets |
|---|---|
| Free | 5 sheets, one-time trial |
| Pro | 50 / month |
| Studio | 200 / month, pooled across seats |
| Enterprise | Unlimited |
Need more than your monthly allowance? Top up at $19 per 100 sheets — top-ups are used after your plan allowance and never expire.
Plans
Billing & limits
One subscription covers all four studios. Global USD, billed through Stripe. Here's what each plan includes; the pricing page has the full comparison.
| Plan | Price | Key allowances |
|---|---|---|
| Free | $0 | 3 mix designs, 1 LCA (watermarked), 2 geo reports / mo · unlimited manual BOQ · 5 extraction sheets one-time · watermarked exports |
| Pro | $79/mo · $64 annual | Unlimited mix & geo · 15 LCA / mo · 50 extraction sheets / mo · clean exports · full history · email support |
| Studio | $199/mo · $159 annual | 3 seats (+$39/seat) · everything in Pro per seat · 200 extraction sheets / mo pooled · shared projects & rates · priority support |
| Enterprise | Custom | SSO, API & webhooks, unlimited extraction, audit logs, dedicated support, on-prem option |
Extraction top-ups are $19 per 100 sheets and never expire. The 14-day Pro trial needs no card. Existing $20/mo Professional subscribers are grandfathered on that rate. Cancel anytime and keep access to the end of the paid period.
Questions
FAQ
Does the AI ever produce the final numbers?
No. Across every studio, quantities, corrections and totals are computed by deterministic code. The AI drafts and extracts; you confirm; the engine calculates. In BOQ specifically, no priced quantity is ever generated from unconfirmed extracted data.
Do I still need a licensed engineer?
Yes. Every output — mix design, LCA report, geotechnical parameters, priced BOQ — is a draft that requires review and sign-off by a licensed engineer. Geotechnical estimates in particular are advisory and don't replace site-specific lab testing.
Can I use BOQ without uploading drawings?
Yes. Manual entry of structural, masonry and roof items is unlimited on every paid plan and never touches your extraction-sheet allowance. Drawing extraction is optional.
Which codes and methods are supported?
Mix Design follows ACI 211, EN 206, TS 500 and BS 8500. BOQ rebar detailing follows ACI 318, EN 1992 or TS 500. LCA uses the EF 3.1 method. Geo applies published SPT/CPT correlations, each named in the output.
What happens to missing data?
It's tracked, not hidden. LCA reports a data-coverage percentage and widens the uncertainty band where inputs are weak; Geo flags inputs outside a correlation's validity range; BOQ turns unclear reads into questions and lists unresolved elements as pending.