the best plan, its reason,
its record.

kint gives teams the best plan the solver finds within their rules, with the gap that says how far from best it can be. Each plan names the limit that decides it and what each resource is worth, and goes on record. Below are the features behind it, from a planner's first sentence to the released plan.

Platform · features
modelsolveanalyserecordconnect
model · solve · analyse · record · connectfive groups behind every plan
In the product

the features, one place to work.

The same features as planners see them: the project at a glance, the agent, search, the rules the plan keeps and the studies to compare.

Loading the product screen

Five groups

the features, by group.

From the data that comes in to the decision that goes out, and the record in between. Each group starts with what planners can count on, then lists the features.

01

model

A planner writes goal and rules in plain language. kint's decision agent drafts the model. In the pilot, kint's forward deployed engineers (FDEs), who work directly with the team, review it with domain experts.

words to a reviewed model

  • agent drafts the model

    a model your planners can read

    Decisions, rules and goal from the planners' description and the data.

  • review in plain languageEvery rule shown as a sentence with a link to where it came from.

  • versioned dataEach upload becomes a version with a content hash. Changes are visible.

  • model reviewIn the pilot, kint's FDEs review the model with domain experts.

  • skill and agent libraryReusable prompts, schemas and agents for decisions that come back.

02

solve

kint picks the method from the structure of the problem: LP, MIP, CP, QP or a search over learned models. For exact models, the solver reports the bound and the gap. At a gap of 0.0 %, the solver has shown that no better plan exists for the model as stated, within its numerical tolerance.

exact and learned

  • how close to optimal

    bound and gap, with each exact plan

    Stored with the run. At a gap of 0.0 %, no better plan exists for the model as stated, within the solver's tolerance.

  • exact solversOpen-source LP, MIP and CP solvers included. Commercial solvers supported, licences arranged per engagement.

  • surrogate optimizationEnsembles of learned models, searched within the stated limits.

  • several goalsWeighted or ordered goals, and the set of best compromises.

03

analyse

Managers' next questions run on the same model and data: which limit binds, what a resource is worth, which strategy holds.

five questions, on the simulation page

  • binding limit

    the limit that decides the outcome

    Named in plain language, with what it would take to move it.

  • sensitivityWhat one more unit of each resource is worth.

  • what-if and scenariosChange inputs and compare strategies.

  • stress testsOne plan against many possible futures, with the limits that break.

  • uncertainty rangesLearned predictions come with an uncertainty interval.

  • explanationsWhich inputs move a learned outcome, and how.

04

record

Each run pins the data version, the model version and the settings. A person on your team releases each plan before it is written back. The record keeps who and when.

accountability built in

  • replay

    any past run, re-checked

    Each run pins the data version, the model version and the solver settings, so the plan can be re-checked against the record.

  • release by your teamAn approved workflow runs whenever your systems call it, and each plan waits for that release.

  • roles and accessRoles separate who may model, run, release and read.

  • EU AI Act documentationDrafts of the technical documentation, kept current from the running system.

05

connect

Data as it is, from databases, APIs and files. Use the kint interface, call the API from existing systems, or embed the engine in a product under your brand.

into your systems

  • file exports and api

    your data as it is, read-only

    Data comes in as file exports (CSV, spreadsheets, JSON) or through an API; further connectors per engagement.

  • workflows and live run viewCompose fetch, analyse, solve and notify into one run. Watch its graph live.

  • api and live progressHTTP API with OpenAPI documentation; live progress over WebSocket.

  • plans outApproved plans go back only through an API you control. The control system decides whether to apply them.

  • oemCall kint through the API and ship it under your brand: kint inside your product.

  • deploymentHosted in Germany and operated by kint. For operations that must not leave your network, we scope an installation inside it.

pick one feature. test it on your data.

Bring one decision and a sample of the data behind it. We run a live session and show you what kint finds, and why.

Models
whitebox · blackbox · graybox · simulation
Industries
manifest · merit · takt · horizon
Controls
security, hosting and the record
Contact
hello@kint.pro