You set the rules,
appto holds them
Who is in the company and on which team, what they can reach, what leaves the building and how much the company spends on AI. One place for the whole organization — and the boundaries you set hold the same in a conversation, in agents and in Slack.
The role grants permissions, the team sets the reach.
Four levels, each one adding to the last. A champion builds the assistants, contexts and skills the rest of the team uses — while money and security stay with the administrator.
- Runs conversations and projects
- Uses the assistants, skills and contexts they were given
- Connects their own accounts to integrations the company allows
- Creates assistants for specific team tasks
- Adds and organizes contexts — documents, procedures, decisions
- Builds skills and keeps their versions in order
- Shares skills and contexts with their teams
- Creates, changes and archives routines with triggers
- Removes people from the organization
- The credit plan, top-ups and the spending ceiling
- Integration policies and the models allowed in the company
- Privacy Gateway and the activity log
Slack is not a back door.Everywhere shows the same.
Permissions are pinned to the person, not to the window they ask from. The same person in appto, in Slack, in Teams and through an agent has the same reach — and when they leave the company, all of it closes at once.
All four columns look identical, and that is the point: these are not four sets of permissions, but four views of one.
Account disabled. All four entrances closed the same day — nobody has to remember them one by one.
Every agent has its model.Nothing switches by itself.
The model is not an accident or a function of server load. The person who builds the agent picks it and that is what the agent runs on — the same every day, for everyone who uses it.


The primary model is chosen by whoever builds the agent, and that is what it runs on every day. The fallback list says what it may switch to mid-conversation — and nothing beyond it, even if the model is available elsewhere in the company.
The bill does not grow in the background.
You buy credits up front and they are the ceiling. When they run out, appto pauses AI work instead of adding to the bill — and if you prefer continuity, you switch on top-ups and set your own limit.
Try it on your own team
You see what caught on.And whether it is growing.
This is the question that comes after three months: is anyone actually using it. You see which skills and tools carry the result — and whether the volume grows week over week or has stalled.
Who, what and when.Without reading conversations.
Logins, roles granted, integrations connected, moves in billing, safeguards switched on — each with an author and a time. The log records the event, not the content of conversations, and entries are written as sentences, not technical names.
Set it up once,for the whole company
Start with one team and one plan. The rest — roles, access, models and the bill — sits in one place and waits until you need it.
Before you bring the company in
The questions owners and security teams ask — about visibility, cost and what happens after someone leaves. The rest is in our trust centre (Polish for now).
Can an administrator read my conversations?
No. The activity log records the event — who did what and when — not the content. Analytics counts usage: how many messages, which skills, which tools. Neither of those shows what you talked about. Email addresses in the log are masked on top of that.
Who can change organization settings?
The administrator and the manager, each within their scope: a manager runs a team and shares skills, contexts and routines with it, while the administrator answers for billing, integrations, allowed models and security. Someone without those roles uses appto but does not change organization settings.
What happens when the credits run out?
appto pauses AI work. The bill does not grow on its own and there is no surprise at the end of the month — the plan is a ceiling, not a suggestion. If you prefer continuity to a hard stop, you switch on automatic top-ups: appto adds another batch of credits when less than a tenth of the plan is left. That is a deliberate decision, not a default.
Can models be limited to part of the company?
Yes. The company decides which models are allowed at all and for whom — three for the whole organization, one for a single team, the full shelf for another. Whoever builds an assistant then picks a model, but only from what is allowed. Nothing switches by itself in the background based on load or cost.
What happens when someone leaves the company?
Removing a person from the organization cuts their access — in appto, in Slack, in Microsoft Teams and in the agents they ran. All of it at once, the same day, because the permissions were pinned to them rather than to a place. The work they left in the company’s projects and contexts stays with the company.
Where do we set which integrations may be connected?
On the tools page — the whole policy board is there: which integration works, for whom, and whether connecting an account is mandatory. Administration shows the wider picture: who actually uses what, on which model, and at what cost.
Does our data go into model training?
No. The request goes to the model provider so an answer can be produced, but your material is not used for training. appto’s infrastructure sits in the European Union, in Ireland. Some model and integration providers operate outside the EU, which is why the subprocessor list is public.