Ts-Knowledge
Sign inCreate an account
All use cases

Use case · Online stores and small businesses

Your whole store, known to your AI — and backed by evidence.

Product pages, terms of sale, return procedures, shipping guide, project meeting notes, technical documentation: Ts-Knowledge brings them together, extracts business objects, rules, processes and decisions, shows where they contradict each other, and serves them to your support agents and your developers alike.

Three people, three answers to the same question

Maison Alcôve has been selling home linens online for six years. Its knowledge is everywhere: a Drive, a website, a GitHub repository, and the heads of two people.

  • “Can I return my duvet cover?”The website says thirty days, the customer-service procedure says fourteen. Nobody noticed — until the first dispute.
  • Why no partial orders for in-store pickup?The decision was made two years ago, and the reasoning sits in meeting notes nobody can find.
  • The AI support agent improvisesPlugged into an incomplete FAQ, it promises delivery to Switzerland — which the store doesn't offer.

The company as a hub

One department, one branch; what they share lives in the hub

The hub holds what the whole company shares: the order, the return, the carrier, the return policy. Each department is a branch, with its own documents, its own permissions, and whatever its documents assert.

  • Your sources, synced: a Google Drive folder, the website through its sitemap, a GitHub repository — every day if you like.
  • Walled off by department: reading a branch only opens the identity of the hub's entities, never the facts of other branches.
  • A contradiction between two departments shows up in the hub — and only there, where both values can be read side by side.
Maison AlcôveOrder, return, shipping, return policy, carriers, ERP…
  • Customer serviceCustomer-service procedures, returns, canned responses — Drive.
  • LogisticsPicking, shipping, delivery zones.
  • Website & termsProduct pages, FAQ, terms of sale — the website, via its sitemap.
  • TechThe store's documentation — GitHub.
Fictional: a home-linen store, four departments, one hub.

1 · Find

The right procedure, cited, for every question

Your teams and agents ask the question the way a customer would. Hybrid search combines meaning and exact wording — a product reference, a carrier name, an error code — and every excerpt arrives with its source.

A procedureHow should a returned parcel that arrives damaged be handled?
Step 3 of the return procedure: photos, no full refund, store credit offered instead.Customer-service procedure — Returns · § 3 · Drive
An exact wordWhich carrier do we use for Corsica?
The zone table in the shipping guide, found by the word “Corsica” as much as by meaning.Shipping guide · Delivery zones
“I don't know”Do you ship to Switzerland?
Not enough evidence: no document says so. Your support agent promises nothing and hands off — instead of inventing a delivery zone.

2 · Understand

The store's objects, rules and decisions

With the “E-commerce” template, Ts-Knowledge reads your documents and extracts what they assert: the order and its states, the Click & Collect feature, the picking process, the pickup rule, the decision and its rationale. Each element comes with the passage that proves it, and nothing is duplicated: a feature points to the objects it touches.

Click & Collect, through its links

  • Touches the Order object, which gains a “ready for pickup” state.
  • Follows the “Prepare an order” process, step by step.
  • Subject to the rule “pickup within 72 hours, otherwise back to stock”.
  • Affected by the decision “no partial orders for pickup” — with its rationale and the option that was ruled out.

Inconsistency · Return policy · deadline

30 days“You have thirty days from receipt…”Terms of sale, article 8 · Website & terms branch · cited
14 days“Returns are accepted within fourteen days.”Customer-service procedure — Returns · Customer service branch · cited

Two citations that disagree: major severity. Both values stay, with their evidence — it's up to you to fix the document that's wrong.

Business objects

Order, cart, payment, shipment: their states, and the transitions that change them.

Processes

The steps of a procedure, who carries them out and with which system — each with its evidence.

Rules and decisions

What the store commits to, and on what conditions; what it decided, why, and what it ruled out.

Inconsistencies

Two incompatible values for the same dimension are flagged and ranked by severity. Neither one overwrites the other.

A new dimension — like this return deadline — expects a single value by default, so two documents that disagree get flagged. The space's metamodel specifies, dimension by dimension, which ones can legitimately have several values.

The Knowledge Snowflake

A map of the store, project after project

Thirteen facets — business objects, features, processes, rules, decisions, offering, identity, organization, partners, systems, incidents, projects, glossary — and one ring per stage: each project's documents, in the order they came. At a glance, you see what each project documented, and what never was.

What the map tells you

  • The glossary is empty: nobody ever wrote down what a “split order” means. The Snowflake doesn't guess — it shows you.
  • Pay in 3 delivered a feature and a partner, but no process: customer service has no procedure for a declined installment plan.
  • The Snowflake graph colors each link by facet: who touches which object, which system stores what, which partner provides which service.
  • Your agent reads it too: get_knowledge_flake gives it the map — where to look, and what's missing — before its first search.
FacetLaunchClick & CollectPay in 3Marketplaces
Business objects
Features
Processes
Rules
Decisions
Offering
Identity
Organization
Partners
Systems
Incidents
Projects
Glossaryempty
Fictional example. The bigger the dot, the more that stage has to say; a dotted circle means it says nothing.

3 · Connect

A support agent that cites, a dev agent that knows what it's touching

Each agent gets its own key, restricted to the spaces that concern it. Claude, ChatGPT, Gemini, Mistral or an IDE agent query Ts-Knowledge over MCP, and write their own answers, with sources to back them.

A customer received her duvet cover twenty days ago and wants to return it. Is that possible?

search_knowledge“return deadline” · Customer service, Website & termsTerms art. 8: 30 days · Customer-service procedure § 2: 14 days
list_knowledge_conflictsMaison AlcôveReturn policy · deadline · 30 d / 14 d · major

The terms of sale allow thirty days Terms of sale, art. 8, the customer-service procedure fourteenCustomer-service procedure, § 2. The two documents contradict each other: I won't pick one — I'm passing this on to the customer-service manager with both passages.

I need to add a state to the order. What depends on it?

get_entity_relationsOrdertouched by: Click & Collect, Pay in 3 · stored by: the ERP
get_knowledge_entityClick & Collectfollows: Prepare an order · subject to: pickup within 72 h
get_entity_evidencepickup within 72 hdocs/orders/pickup.md · Tech

Two features touch the order. Click & Collect follows the picking process and the 72-hour rule docs/orders/pickup.md: make sure the new state doesn't bypass it.

Ts-Knowledge doesn't write anything and calls no model to answer: it provides the material and the sources, and the agent you chose writes the answer.

Get started in four steps

  1. Create your departments

    One space per department, grouped under a hub — the company — in one click.

  2. Connect your sources

    A Drive folder, the website's sitemap, a GitHub repository, or plain files. Manual or daily sync.

  3. Turn on knowledge

    Choose the “E-commerce” template: objects, rules, processes, decisions and the Snowflake build up with every import.

  4. One key per agent

    Support reads Customer service and the terms of sale, dev reads Tech. Each agent has its own key, and its own permissions.

  • Google Drive
  • Website and sitemap
  • GitHub
  • Claude, ChatGPT, Gemini, Mistral

Other use cases

Series authorsYour series bible, kept up to date on its ownBooks, drafts and notes from the same universe: who knows what, what contradicts what, what is still missing — without ever spoiling what comes next.See the use case Law firms and legal departmentsEvery matter walled off, the law as of the right dateLetters, expert reports and court decisions split along their natural structure, French law as a shared library, and a timeline for every matter.See the use case Software companies and product teamsYour docs and code, queryable from the IDEA synced GitHub repository, its services and its architecture decisions, within reach of every developer's agent.See the use case

A single source of truth for the whole store

Create your departments, connect your Drive and your website, and ask Ts-Knowledge the question your customers ask every day.

Create an accountTake the product tour