An AI agent in the team chat
Built for a relocation company with teams across India
An AI agent sits in the team's group chat, answers from the company's own knowledge and takes on requests.
Built
We run our own company on AI agents, and we build the same for others. Here are 22 pieces chosen for range: 15 kinds of work, in 9 industries and teams, on 10 of the 12 layers a company runs on. Each says what kind of company it was built for, never who.
By industry and kind of work, never by a client's name. Open any one for what it does, the kind of company it was built for and how its number was counted.
52 public systems we have built and can show, in 18 kinds of work, counted on 12 October 2026. These 22 were chosen for range, and every one of the 52 is in the index at the foot of this page.
Six pieces, each with the kind of organisation it was built for.
Built for a relocation company with teams across India
An AI agent sits in the team's group chat, answers from the company's own knowledge and takes on requests.
Built
Built for an industry body in Asia Pacific, for its leadership meetings
Delegates carry the whole day on their phone, in Japanese or English at a tap: every session, live captions, and a written recap of every session the same day.
Live
Built for a relocation company with teams across India
Built so a relocation company can run on one system, with every case, vendor, payment, visa and home in it.
Live behind a sign in
Built for the sales team of a relocation company in India
It finds the right people at a target company from its own public pages and filings, records the source of every field, and it cannot send a single email.
Built
Built for a relocation company with teams across India
Everything the company knows sits on one map you can fly through and search in a second.
16,350 pieces of company knowledge on one map, searched in a second. Build of 11 Oct 2026.
Live behind a sign in
3 bed flat, AundhPark facing
2 bed flat, BanerNear the school
3 bed villa, WakadGarden, two parking
3 bed flat, Kalyani NagarHigh floorBuilt for a relocation company with teams across India, for the families it moves
After a day of house hunting, the consultant's spreadsheet turns into one page on the family's phone.
128 photographs of 14 homes on one phone page, from its first tour's spreadsheet. First tour, 8 and 9 Jul 2026.
Built
16 systems our own company runs on, from the AI agents that do the work to the guards that watch them.
Our own GPU machine runs six kinds of AI work: it reasons, reads photos, makes pictures and music, listens and speaks.
6 kinds of AI work on one private machine: reasoning, seeing, pictures, music, speech in and speech out. Proved 22 Sep 2026.
Built
Seven connectors give our AI agents safe doors into mail, calendar, tasks, documents, meetings, saved research and memory.
161 tools across seven agent connectors, counted from source on 12 Oct 2026.
Built
Built for our own company, for every automation on its machines
Every night, with no AI model, it checks every scheduled job, AI agent tool, rule and memory on every machine it can reach.
635 scheduled jobs, agent tools, rules and memories checked on the night of 11 Oct 2026, with no AI model.
Runs every night

Our brand films are made in code, with every frame drawn by software.
31 second brand film with every frame drawn in code. Made 9 Oct 2026.
Built
Built for our founder's own products, in June 2026, before Agentec
One set of design rules makes an analytics app, an ERP, a live operations screen and a phone app look made by one studio.
4 kinds of product dressed by one set of design rules, from an ERP to a phone app, June 2026.
Built
Built for our founder's own work, in June 2026, before Agentec
Over twenty nights, a team of five AI agents worked while our founder slept: one on open projects, one turning the day's mistakes into lessons, one scouting new tools and one weighing new AI models against ours.
68 overnight reports filed across 20 nights by a five agent team. 15 Jun to 8 Jul 2026.
Ran 15 Jun to 8 Jul 2026
Our own brand is built like software.
22 logo files built from one recipe. On 9 Oct 2026.
Built
A machine running many AI agents at once needs guards.
9 guards watching the machine the agents run on: memory, disk, processor, the kernel and stuck processes, counted 12 Oct 2026.
Runs all day
Built for our own company, over every AI agent it runs
One harness over every AI agent we run: what each may load and touch, a check before each action, and a record of what actually happened.
Built
Each trade page has an AI that speaks that trade.
Live
Built for our own company and a relocation company with teams across India, for the products they launch
Product films made from the product itself.
4 launch films for one product, cut in code. Read 11 Oct 2026.
Built for our own company, for every AI service it pays for
Every AI service the company pays for in one place: how much of each limit is left, when it resets, how long it will last at today's pace, and every bill.
Built for our own company, on a private GPU machine of its own
One AI agent on a private GPU machine of its own, with its models and its voice on that machine, reached by chat, a phone app, a desktop app and two computers, all sharing one memory.
5 ways in to one agent with one memory: chat, a phone app, a desktop app and two computers. Read 11 Oct 2026.
Built for relocation firms that want a family portal under their own name, shown with made-up brands
One tap puts a different firm's name and colours on the whole product, the family's portal, the city guide and the concierge, with nothing rebuilt, and the portal reshapes itself for each client's policy and each family.
Built for our own company, for every agent at work on its machines
Every AI agent at work, what each is doing this minute, and one queue for the moments they need a person, in their own words with one button each, built on the alarm rules of industrial control rooms.
A model of how our founder talks, built from more than 600,000 of his own dictated words, scores any draft and flags the lines that sound least like him.
93 against 56 the founder's own words against an AI agent's drafts on the same five subjects, scored on 22 Sep 2026.
Runs every night
Nothing in the selected work is both.
Clear one of the two filters, or search everything we have built below.
We never name a client or show how they run inside. Each piece says what kind of company it was built for. Where a real screen would show someone's data, we redraw it with sample data and say so.
52 public systems in 18 kinds of work, in the order of the layers a company runs on. Search them, or open any one for what it does.
52 systems
AI agents that take whole jobs off a team and bring a person every action that matters, one tap each.
Several AI agents on one job, each with its own part, working through the night if the job needs it.
Safe doors for AI agents into mail, calendars, tasks, records and documents, with every outside send held for a person.
Speaking instead of typing: words that land as you talk, and companions you can talk to.
Every message read and sorted, and every reply drafted for a person to send.
Engines that find the right buyers, do the homework on each one and keep every lead warm.
Everything a company knows or said in a meeting, found in seconds, with the source beside every answer.
Reports, quotes, invoices and records, drafted, checked and signed in the company's own look.
Layers that read and judge for you: a market watched, numbers explained, a draft scored against your own voice.
The one system a company runs on, with every case, customer, supplier and payment in it, and a view for each person who uses it.
Apps for customers, portals for partners and tools for the team, each person signed in to their own view.
Websites and guide engines that people, search engines and AI assistants can all read and use.
Brands and design rules built as code, so every screen and file looks made by one studio.
Brand and product films made in code and checked frame by frame before they go out.
A new business taken from an idea to research, a brand, decks, films, a website and a working demo.
AI models that run on a machine the company owns, so the work and the data stay with it.
The rules around every AI agent: what it may load and touch, a check before each action, and a record of what it did.
Audits, guards and control rooms that show what every AI system did and stop what it should not do.
Nothing by that name, yet.
Most of what we build started as a problem nobody had named. Tell us what you need.
45 minutes on a video call with our founder and the people who live with the problem. Nothing to sign. You leave knowing what we would build first, and why.
Or write to us.hello@agentec.cloudAgentec on LinkedIn
An AI answers here and can be wrong. Nothing is promised until a person writes it.
Tell me what slows your company down.A real answer in a few seconds, with the system we would build beside it. No name, no email.