CodeTrain / Docs / Running a team

Running a team on CodeTrain

This page covers the Team plan from both ends. If you bought it or you are deciding, it walks through creating a team, inviting people, paying for AI usage, team keys, onboarding programs and the evidence pack. If you were invited to a team, read "What a lead sees" to know exactly what your lead can see about you and what they never see.

Set up a team

Buy the Team plan on the dashboard's "Plan & billing" tab. Pick a team size ("Up to 5", "Up to 25" or "Up to 100" learners), pick a Usage Bundle (1x, 2x or 4x per learner) or none, then pay by card in checkout. Then go to the "Devices & Team" tab, type a name and press "Create team". One owner can run more than one team.

To add a person, the owner or an admin enters their email address, chooses admin or member as the role, and says whether that person is a learner. CodeTrain emails them an invite. Someone with no CodeTrain account yet gets one when they first sign in with that email address. The invite mail and the dashboard both tell the person what the team's owners and admins can see about them.

Who can do what

The owner is the person who bought the plan. The owner and admins manage a team together: they add and remove people, mark learners, choose how usage is shared, add team keys, set up onboarding programs, and export the evidence pack. Only the owner buys usage, changes or cancels the plan, and deletes a team. The owner cannot be removed from a team. A member who is neither owner nor admin manages nothing; they see their own row through "What your lead sees".

Learners and the team size

The team size counts learners only. Anyone added as a non-learner is always let in and does not count against it. A learner already marked on another of the owner's teams costs nothing extra, and their usage stays with the team where they were marked first. When the size is full, another learner cannot be added until somebody is marked out or the owner picks a larger size. Inviting someone as a non-learner still works. A non-learner cannot start a lesson on the team, but they can see the team's screens.

Prices

The table below shows a month's total for each team size with each choice of bundle. The flat fee buys the software; it holds no AI usage.

A month
LearnersOwn model key1x bundle2x bundle4x bundle
Up to 5$79$164$234$374
Up to 25$149$514$814$1,414
Up to 100$399$1,664$2,814$5,114

Usage Bundles

A team that runs on its own model key needs no bundle. A Usage Bundle adds managed AI, meaning AI that CodeTrain runs for the team. One unit is the same usage as one Pro learner gets in a month. "Add usage" buys units for the current billing period only, charged at once, and the recurring bundle does not change. "Auto-raise" is off by default. When it is on and the team has used 90% of the period's usage, it buys Add usage on its own up to a ceiling the owner sets for each billing period, and the owner is emailed each time. It never changes next month's bill.

How a team shares its usage

Each team draws its own part of the owner's Usage Bundle, in proportion to its learners. The owner and the team's admins choose how that team's learners share its part. "Even split" divides the part equally between the learners. "Shared pool" lets the learners draw from the part until it is gone. The third choice runs an even split until a day of the billing period the lead sets, then the shared pool until the period ends. The choice can be changed at any time and never changes what is paid. A daily limit per learner stays on in every choice, so one runaway loop cannot use the part up.

Team keys

Team keys let a team run tutoring on its own model account. The providers offered are Anthropic, OpenRouter, Amazon Bedrock, Claude models on Google Cloud, Google Gemini with an API key, Ollama (self-hosted), Ollama Cloud, and any OpenAI-compatible endpoint. A key is stored encrypted and is never shown again; "Test" sends a one-token request with it. Members use the team's default key unless a manager assigns them a particular one. If a key fails, lessons fall back to usage CodeTrain manages, unless the team turns on "Require BYO-key", which blocks that fallback. A second box still allows managed usage for lessons taken in the browser.

What a lead sees

For each member, the owners and admins see per team: the member's email address and role, lessons started and completed, "Unaided" (steps written correctly the first time with no hint, no skip and no code written by the learner's AI assistant), how many of those steps were beginner, intermediate or advanced, "Needed a retry", "Relapsed concepts" (concepts that came back as a struggle within 14 days of being passed), "Mastery lost", concepts, "Shaky" (concepts due for review today, then those due later), skips, usage as a percent of the member's budget or share, the number of tokens the member ran on their own model key, and the headings of steps the member keeps struggling on.

If you were invited to a team, this is the whole list. Owners and admins never see your code, your questions to the tutor or the tutor's feedback. The one exception is a new hire on an onboarding program, whose first task shows a lead a summary and part of its code diff. You see the same row through "What your lead sees" on the "Devices & Team" tab. There is no way to hide the row or opt out of it.

Onboarding programs

$ codetrain onboarding generate
$ codetrain start --mode onboarding

Onboarding programs are in beta. A team can have several named programs. For each, a lead fills in up to seven sections: architecture, local setup, coding standards, security, incident response, a first task, and check-in cadence, by hand or with "Draft with AI". The lead then runs codetrain onboarding generate in the repository, where --program names a program, and it writes the journey: classes on the codebase, the team's practices, and a first task. A new hire runs codetrain start --mode onboarding in the repository, with --program to pick one when there are several. The lead sees each hire's status, position, lessons completed, the time from joining the program to the first task, and the first task's summary. The first task is a suggested change shown as a diff, never written into the repository for the hire.

Delivery numbers

$ codetrain metrics --upload --team NAME_OR_ID

codetrain metrics --upload sends one repository's delivery counts to a team. It needs a login as a team owner or admin, runs once a day per repository, and sends counts only, no code. A lead who manages several teams names one with --team NAME_OR_ID. The team's "Delivery in your repositories" card shows the figures per repository. A lead can type the team's monthly AI tool spend, which CodeTrain cannot see, for the figures worked out per dollar; owners and admins see the amount, and every member sees the figures worked out from it.

See also: Bringing the numbers into CodeTrain

The evidence pack

"The team's evidence pack for audits" downloads, as CSV or JSON, each member's learning figures for the recent window and the team's delivery totals. It holds counts and percentages only, with no learner code, questions or tutor feedback. Owners and admins export it. Each member sees the date and format of every export made while they were on the team. The pack can support records for ISO 27001 clause 7.2 and EU AI Act Article 4.

Change or cancel the plan

The owner changes the size or the bundle on "Plan & billing". The size must hold every marked learner, so learners are marked out first to pick a smaller size. A smaller bundle starts at renewal; the plan keeps what it holds until then and nothing is refunded. Cancelling keeps the size and bundle until the end of the period already paid for, then the plan ends. Auto-raise is turned off, and nothing is refunded or charged.

Delete a team

Only the owner deletes a team, by typing its exact name. Onboarding programs can be moved to another team the same owner runs first; everything else in the team, its members and team keys included, is removed. A person's personal workspace cannot be deleted this way.