STEP 1
Paste your list
Drop attendee rows in from any spreadsheet or CSV. A name column is enough; more columns make the teams smarter.
Paste a list. Get teams built on purpose.
Set 8 dimensions from similar to diverse, mark who stays together and who stays apart, and a patented engine (US 10,007,721 and US 11,487,802) builds every round right in your browser. Nothing gets uploaded.
1. Load your people
Paste rows from any spreadsheet (Excel or Google Sheets: select, copy, paste) or choose a CSV export from Eventbrite, Luma, Cvent, Zoom, or anywhere else. A name column is all that is required; organization, title, expertise, asks, offers, geography, and who-knows-whom make the teams smarter. Extra columns become sliders automatically.
Or start from a scenario, teams appear instantly:
Your list never leaves this tab. Parsing, scoring, and team formation all run in your browser. Nothing is uploaded, stored, or seen by us. Close the tab and it is gone.
Why this beats pasting your list into an AI chat: no one is dropped or duplicated, team sizes are exact, your keep-together and keep-apart rules are enforced or reported when they cannot be, and the same seed always reproduces the same teams.
Every person placed once. Team sizes exact. Every keep-together and keep-apart rule enforced, or reported plainly when it cannot hold. Same seed, same teams, every time.
TeamCreator by United We Transform
Your teams, built on purpose
Teams built free at unitedwetransform.com
How TeamCreator works
Four steps, no login, and nothing you paste leaves your browser. The browser engine is a faithful JavaScript port of the Python reference in the GitHub repo; both return byte-identical team assignments on the same inputs, and the hosted MCP tool runs that identical JavaScript.
STEP 2
Set the 8 dimensions
Slide each of the 8 dimensions from similar to diverse. You decide, dimension by dimension, where like should group with like and where the room should mix.
STEP 3
Add keep-together and keep-apart rules
Name who must share a team and who must not. TeamCreator enforces each rule, or reports plainly when one cannot hold.
STEP 4
Get multi-round rotation
Ask for several rounds and the engine minimizes repeat pairings, so people meet new faces each round instead of the same group twice.
Similar vs diverse, dimension by dimension
A good table depends on the job. Peer learning often needs shared context. Innovation work needs complementary skills. Networking needs useful weak ties and fewer obvious repeats. So TeamCreator does not give you one similar-or-diverse dial. It gives you 8, one per dimension, and you set each on purpose.
Expertise
Similar groups deep specialists together; diverse mixes disciplines
Organization
Similar keeps colleagues together; diverse breaks up company clusters
Geography
Similar groups by location; diverse mixes regions
Seniority
Similar makes peer groups; diverse mixes levels
Gender
Similar groups by stated gender; diverse balances
Familiarity
Similar seats people who know each other; diverse builds new ties
Ask and offer fit
Similar pairs people whose offers answer each other's asks
Helping potential
Weights pairs where one person can concretely help the other
This is the phrase to remember: similar vs diverse, dimension by dimension. Academic team-formation tools such as CATME already validated the demand for exactly this kind of control. TeamCreator puts it in your browser, free, with the rules enforced and the result reproducible.
How TeamCreator compares
Most people build teams one of three ways: paste the list into an AI chatbot, wrestle a spreadsheet, or use whatever an event app bundles in. Each does something well. Only TeamCreator runs one deterministic, patented method that gets the math right every run, enforces and reports your rules, and never leaves your browser.
Why pasting your list into ChatGPT goes wrong
An AI chatbot has no calculator behind it. It predicts a plausible answer one token at a time, so it can miscount a team size, quietly drop or duplicate a name, and tell you it followed your keep-apart rules when it did not. It has no fixed balancing method unless you hand it the exact dimensions, weights, and data structure yourself, which almost no one does. And to use it at all you paste your attendee list, names and all, into someone else's servers. TeamCreator runs the same method every time, in your browser, and shows you every rule it kept or broke.
| Capability | TeamCreator | AI chatbot (ChatGPT, Gemini) | Spreadsheet (Excel, Sheets) | Event app |
|---|---|---|---|---|
| Gets the math right every time | Deterministic code, exact counts every run | Predicts an answer, so it can miscount, drop or duplicate people | Correct, but you build every formula | Correct, but rigid |
| Same list in, same teams out | Yes, same list, settings, and seed | No, answers differently each run | Yes, until a cell changes | Usually re-randomizes |
| Honors keep-together and keep-apart rules | Enforced, and tells you exactly what it had to break | Says it will, then can break them silently | Manual and error-prone | Limited, and no report of what broke |
| Balances on 8 dimensions you control | One slider each | Only if you hand it the exact method, weights, and structure | Complex manual formulas | Fixed categories |
| Multi-round rotation without repeat pairings | Yes, minimizes repeats across rounds | Loses track across rounds | Very hard to build | Rare |
| Uses any data you already have on people | Paste any columns you have | Yes, but retyped with no method behind it | Yes, but manual | No, locked to the app's own fields |
| Your list never leaves your device | Runs in your browser, nothing uploaded | You paste attendee names into a third party that may train on them | Local to your file | Uploaded to their servers, account required |
| Free, no account, no usage wall | Free, no account | Free and paid tiers | Free | Free trial, then paid |
| Patented method | Patented (US 10,007,721 and US 11,487,802), built by the patent holder | No | No | Replicating outcome-based grouping may require a license |
Random grouping vs TeamCreator
Here is a test you can run yourself right now, no faith required. Paste the sample list, add two or three keep-apart rules, and press Create the teams. TeamCreator either honors every rule or tells you exactly which one it had to break and why.
Now imagine drawing the same groups from a hat. A random draw does not know your rules, so it breaks them at a rate that climbs with every constraint you add, and it re-shuffles into different teams on every run, so you can never reproduce or defend a result. TeamCreator is deterministic: the same list, the same settings, and the same seed always return the same teams. That is what makes it explainable, and it is why a mechanical demonstration beats a self-declared superiority claim.
See it from the founder
Founder video: coming soon
A short founder walkthrough of the method will live here, directly below the tool. This is a deliberate placeholder until the real clip is published. We do not fake a demo.
What early users say
Beta testimonial: pending
Real quotes from early users will appear here once they are collected and confirmed. This slot stays empty on purpose. We never publish an invented testimonial or a usage count we cannot back.
Why patented matters
The method behind the teams is patented
TeamCreator runs a patented team-formation method owned by Collaboration.Ai: US 10,007,721 (a method for reducing human bias when splitting large groups into teams) and US 11,487,802. This is not a language model's best guess at a seating chart. It is a verified, deterministic method, version 0.10.0. The GitHub repo is the Python reference; the browser engine on this page is a faithful JavaScript port of it, and the two return byte-identical team assignments on the same inputs, verified against shared parity vectors. The page and the hosted MCP tool run that identical JavaScript: same input, same teams, every time.
The code is free for noncommercial use under PolyForm Noncommercial 1.0.0. Event platforms and other commercial users can license the method from Collaboration.Ai at collaboration.ai.
Questions people ask
Is it really free?
Yes. TeamCreator is free with no account and no paywall. Not five free tries then a wall. Every feature, every round, and every export stays free.
Does my list ever leave my browser?
No. TeamCreator runs fully in your browser. The names you paste are processed on your device and are never uploaded to a server.
What are keep-together and keep-apart?
Rules you set. Keep-together always places named people on the same team; keep-apart never does. TeamCreator reports plainly if any rule had to break.
What is multi-round rotation?
It builds several rounds of teams from one list while minimizing repeat pairings, so people meet new faces each round instead of the same group twice.
Is it really patented?
Yes. The method is covered by US patents 10,007,721 and 11,487,802 (Collaboration Ai), including a method for reducing human bias when splitting large groups into teams.
What does similar vs diverse mean?
You set 8 dimensions on a slider from similar to diverse. Similar groups like with like; diverse mixes them. You choose the setting for each dimension.
Run it yourself, keep your data local
The open, noncommercial-licensed code runs with Claude, ChatGPT, Codex, or Gemini on your own machine. Tell your agent to pull attendees from your event platform and it wires the connection itself: Eventbrite by API token, everything else by CSV export. Multi-round rotations, previous-team avoidance, and survey design included.
GitHub
The open noncommercial code lives at github.com/Collaboration-ai/teamcreator-uwt. Clone it, install it, and run the test suite yourself.
git clone https://github.com/Collaboration-ai/teamcreator-uwt && cd teamcreator-uwt && pip install -e . && pytest
With your AI
Clone the repo and point Claude Code, Codex, or any agent at AGENTS.md. It reads the file and wires the engine into your workflow: roster in, teams out, nothing leaves your machine. Or skip the clone and call the hosted MCP tool make_teams directly.
TeamCreator use-case pages
Start from the specific job: conference networking, workshop breakouts, innovation sprint teams, Eventbrite exports, Google Sheets, sponsor matchmaking, peer circles, or private local team formation.