Platform Architecture
The architecture that turns owned data into owned task intelligence.
KynticAI sits above the systems a company already owns. Scout proves the Universal Context Layer for free, Fortress turns it into a private Rust/LanceDB runtime, and Elite adds the on-prem model path so the best next task can be explained without giving up the customer data plane.
Buyer spark
This is the page where the buyer should finally say: I get it.
KynticAI is not another model wrapper. It is the missing relationship layer above company data: inject the items, preserve the attribution path, compare similar outcomes, and give the model JSON it can actually explain.
Scout proves the layer for free, in PostgreSQL/pgvector, where the buyer can inspect it.
Fortress moves the same story into the proprietary Rust/LanceDB runtime for serious private scale.
Elite adds the on-prem model path so the explanation can stay sovereign too.
In plain English
Why buyers lean in: they do not have to replace the estate to unlock the relationships.
Free open-source Scout creates the private relationship memory
Scout turns exact authorised data items into attribution paths, relationship sets, PostgreSQL/pgvector proof storage, top examples, and JSON that Fortress sends to the customer's LLM or Elite sends to the KynticAI open-source on-prem LLM model.
Enterprise makes the analysis fast and private
The proprietary Rust/LanceDB engine weights, traverses, and compares millions of relationship sets so private evidence can explain conversion probability, churn pressure, support drag, operational risk, and stale opportunities.
Outcomes make it compound
Approved conversions, losses, saves, support resolutions, and task results feed the relationship layer so future recommendations become sharper over time.
Step-by-step example
A simple enquiry becomes a relationship-backed workflow.
Privacy-safe synthetic example backed by a real validation path. The example shows how the architecture works before a customer walkthrough replaces the sample data with authorised sources and agreed measures.
01 / Live enquiry
A buyer signal lands in scattered systems
email = testname@test.com
cookie = web_cookie_4281
web_search = page_a
product_interest = product_b
crm_status = new enquiry
Free open-source Scout injects the data items and stores the attribution path without moving raw records into KynticAI product operations.
02 / Rust analysis
Similar relationship sets become weight
attributionPath = email -> page_a -> product_b
similarConverted = email + page_a + product_b
similarNotConverted = email_only
confidence = 0.78
sourceTrail = email + web + product + outcome
Enterprise/Fortress adds proprietary Rust/LanceDB relationship weighting and traversal inside the customer estate.
03 / JSON task brief
The top examples rank the next move
topExample = previous converted path
option_1 = follow_up_email | estimated_probability = 72%
option_2 = account_registration_prompt | estimated_probability = 61%
handoff = JSON to customer LLM
Fortress lets the customer's chosen LLM explain the task; Elite can use the KynticAI on-prem open-source model instead.
Product and operations architecture map
Each public product and operating surface has a different job.
Scout is the free, open-source Universal Context Layer (UCL) entry point, and it is separate from KynticAI product operations. Clarity is not the Universal Context Layer, and Importance is not required for it to operate. KynticAI Limited is the legal entity. The separation keeps customer data ownership, commercial operations, and optional AI workflow products clear.
Free open-source data plane
Universal Context Layer (UCL) / Scout
Scout is the free, open-source Universal Context Layer entry point: it injects authorised data items, stores relationship sets, attribution paths, and vectors in PostgreSQL/pgvector, and exposes governed top-example JSON inside the customer-owned data plane for proof and lower-load use.
Canonical Rust engine
Enterprise
Enterprise contains the proprietary Rust relationship, weighting, traversal, and LanceDB vector-analysis engine used to compare millions of relationship sets before Fortress sends JSON to the customer's LLM or Elite uses KynticAI's open-source on-prem LLM model.
Product operations only
KynticAI Operations Layer
KynticAI uses its operations layer to manage accounts, licences, downloads, support, update metadata, registration, aggregate usage, audit, and health. It is not sold as the customer data product.
Separate optional product
Importance Engine
Can optionally consume governed relationship output, conversation signals, and model-feedback signals to rank decisions, positive-response agents, and forensic pattern matches, but it is not a required part of Universal Context Layer (UCL) / Scout.
Separate optional product
Clarity Engine
Can optionally clarify prompts before evidence or model routing, but Universal Context Layer (UCL) / Scout does not require Clarity to operate.
Operating flow
The core Universal Context Layer path, with optional engines around it
01
Free open-source Scout injects authorised data items
Scout is the Universal Context Layer (UCL) entry point: connectors and approved imports load data items, attribution paths, relationship facts, provenance, and APIs inside the customer-owned data plane.
02
Enterprise compares relationship sets
Enterprise/Fortress adds the sovereign runtime, LanceDB vector database, and proprietary Rust relationship, weighting, traversal, and similarity engine.
03
JSON feeds the right LLM path
Fortress supplies top-example JSON to the customer's chosen LLM, such as ChatGPT Enterprise or an internal model. Elite adds KynticAI's open-source on-prem LLM model so the task brief is generated without third-party token charges.
04
Optional products can plug in
Clarity Engine can clarify intent first, and Importance Engine can add decision weighting, positive-response scoring, or forensic pattern matching, but neither is required for Universal Context Layer (UCL) / Scout to operate.
Control model
The control model keeps customer data and commercial metadata separate.
KynticAI operations stay separate
Accounts, licences, support, downloads, update posture, data-plane registration, audit, and aggregate health are product operations only.
Private runtime keeps control close
Connector credentials, source records, relationship facts, and private configuration stay close to the systems that own the work.
Optional products stay separate
Clarity and Importance may integrate with Universal Context Layer (UCL) / Scout or Enterprise/Fortress when useful, but they are separate products with their own routes.
KynticAI Limited is the legal entity
The public brand is KynticAI; the legal entity behind the website and commercial conversations is KynticAI Limited.