Platform
One system for the whole brand.
Customers, purchasing, retail and reporting run from one place, on your data, with your approval. Nothing to learn, nothing to configure.
Team
The roles that run your operations.
Four roles cover the work that eats a founder's week. Each one has a job description, a signing limit and a handbook, like anyone you would hire.
Customer desk
Every inquiry answered within the hour. Buying signals flagged. Nothing buried in the inbox.
Purchasing
"Order 12 more of ref. 1092 steel before October. Last Q4 you sold out by week 3." Seasonality included.
Retail density
Where your tier is carried, in which market, at what density. The market picture before the first call.
Reporting
Monday, 7 am: sales, stock, open orders, overdue VIPs. One page, same format every week.
Knowledge
Four layers. One system.
Strategy tells the AI what the brand is. SOPs tell your people how to work with it. Skills tell the AI how the work is done. Chat is where you ask.
Strategy
The AI knows the brand. Positioning, priorities, markets, price rules. Written once. Every draft follows it.
SOP
Your people know the drill. One procedure per job. Load it in chat, follow it, hand it to the next person.
Skills
The AI knows how. Named, tested procedures it runs on its own: draft the reorder, answer the inquiry, build the report. Versioned like code.
Chat
You just ask. Web, Telegram, Claude or ChatGPT. Plain language in, finished work out.
In a small company that knowledge lives in one head. Here it lives in the company.
Approval
Your name is on the dial. Nothing ships without you.
No mail goes out, no franc is spent, no promise is made to a customer until you have seen it. One tap on your phone: approve, change, stop.
Every outgoing mail waits for your tap
Signing limits per role, in francs, like a real employee
Full logbook, always visible
Missions
Hand over the project. Keep the last word.
"Launch the steel GMT in September" is not a task. It is a project. It gets broken into pieces, worked in parallel, reviewed, and comes back to you as one thing: the finished result, waiting for your go.
Multi-step execution
Long missions with built-in review loops. Nothing merges until you say go.
Your standard, every time
Strategy and handbook live in the system. Versioned, current, followed on every mission.
Finished work, not a longer list
Hand over the mission. Get back the result, not more to-dos.
Watch Market Intelligence
The industry, mapped and comparable.
Ask about one brand, five competitors, a country or the whole market. The answer comes from tracked signals across organic, paid, reviews and retail density, with coverage stated first and every number carrying its confidence.
Brand Card
One brand, one card: coverage, growth, paid, launches, reviews, retail density, strategy and the one open flank. Each line dated and sourced.
Competitive Pack
Two to six brands side by side on aligned signals. A missing signal is read as untracked, not weak.
Market Scan
Which brands are accelerating, which show warning signs, which just changed direction. Grouped into bands across the whole field.
Trajectory
30, 60 and 90 day growth windows and the shape of the curve, so a single good month is never called acceleration.
Where the Ads Land
EU reach by country, reach band against the market, funnel type, format mix, seasonality, and the brands whose ads actually overlap yours.
Launch Style
Organic rockets, paid amplified drops and silent launches over the last twelve months, and whether paid runs before or after organic.
Review Sentiment
Reviewer sentiment, recommendation share and price perception from earned coverage, with the top review linked.
Audience Voice
What collectors say back: themes and sentiment split, weighted by reaction. As counts, never as names.
Market Density
Paid media competitive density per country and month, read against the industry pulse of launches, sentiment and trend clusters.
Coverage stated first. Bands, not fake ranks. Estimates stay estimates.
Works with Claude and ChatGPT
Ask your business. Get factual answers.
Open Claude or ChatGPT and talk to your company. Which orders are open? Draft the reorder for the steel GMTs. Who bought twice this year? Real numbers from your real data. It proposes. You release.
Your company inside the AI you already pay for
Assign work from a chat window
MCP and CLI for developers at agents.concierca.ch
Native data
Your data lives in one place. Your team reads the short version.
Storefront, inbox, orders, suppliers and spreadsheets are imported into your own database in Switzerland. Not fetched from five tools on every question, stored once. On top sit views: reorder candidates, overdue VIPs, open orders, this week's numbers. An agent reads a view, not a hundred documents.
Imported, not fetched
Your systems are synced into one Postgres database in Zurich. Every table belongs to your brand alone. Nothing is pulled live from a third-party API while you wait.
Views, not documents
SQL views condense the raw tables into the questions that matter: what sold, what runs out, who went quiet. The agent gets the answer table, already reduced.
Small on purpose
Less data per question means faster answers, fewer tokens, lower cost and fewer places to be wrong. The same design runs the Monday report and a chat reply.
Embedded for meaning
Knowledge, memories and documents carry vector embeddings in the same database. Search runs on meaning, so the team finds what you meant, not just what you typed.
API and MCP
Everything in the app, reachable by an agent.
Claude, ChatGPT, Cursor or your own code connect over MCP and run as a named user with that user's role. Every call is logged. Nothing gated runs without a person.
Missions
List, validate, run, pause, resume. Read the trace of any run: steps, receipts, cost, tokens. Compare versions and runs.
Knowledge and memory
Search and read strategy, SOPs, skills and memories. Save new ones. Visibility follows the company, department and item rules.
Data views
Query the views your company exposes: orders, stock, customers, runs. Create and update rows where the role allows it.
Approvals and inbox
List what is waiting for a human, reply to inbox threads, resolve an approval. Preview first, then confirm. No silent writes.
Agents and skills
Create, update and run agents. Manage schedules. Read run logs and stats. Skills are versioned and callable by name.
Watch Market Intelligence
Brand cards, comparisons, benchmarks, market density and the signal feed, as tools. Coverage stated first, estimates stay estimates.
Models
You are not betting on one AI company.
Your system runs on every major model. Claude, GPT, Gemini, and fast open ones like Kimi and GLM. Routine work runs on cheap models, judgment on the strong ones, and you see what each costs.
When a better model appears, you get it. No migration, no rebuild, no waiting for a vendor to catch up. The models change every few months. Your company should not have to.
Built on OpenRouter. Every major provider, one bill, no lock-in.
