# Keep your projects and their data on servers you control

> Akollo runs in our cloud, in your own cloud account or in your own data centre. It is the same product in all three. You decide where the data lives, what may leave the network and where the AI runs.

- [Book a demo](https://akollo.com/en/demo)
- [Read the docs](https://docs.akollo.com/en)

## From the first call to a live installation

The installation is the same product as the cloud. These are the decisions your infrastructure and security teams make on the way.

1. **Choose where Akollo runs** (Deployment) — Our cloud, a dedicated environment in your own cloud account, or servers in your data centre. Your teams plan and track work the same way in all three.
2. **Size the machines for your headcount** (Installation guide) — The guide sizes the web application, background jobs, database, backups and monitoring for a pilot, 1,000 or 5,000 people. Without an AI model, no graphics card is needed.
3. **Close the network, then open only what you need** (Security settings) — Connectors such as directory sync reach only the hosts your administrators allow. With no allowed list, they reach nothing at all.
4. **Decide where the AI runs** (AI connections) — Connect your own model server or provider key, use Akollo's models, or switch AI off. Data classified as restricted never goes to an outside model.
5. **Set retention and keep the record** (Governance) — One retention rule per kind of data, and an audit log of every privilege and configuration change for your security team.

## What your teams and administrators work in

Project teams get the full product. Administrators get the screens that keep the installation inside your rules.

### The same project board

Projects, tasks, timesheets and plans work exactly as in the cloud. People on an installation do not get a cut-down version.

- Boards, timelines and lists
- Time tracking and approvals
- English, Turkish and German

### Audit records

Changes to privileges and configuration, with who made them and when, for administrators and your security team.

- Role and setting changes
- Who, what and when
- Read under Admin

### AI connection

Add your own provider key or model server, choose which level of data may go through it, then test it before it goes live.

- Key stored encrypted, never shown again
- Data level per connection
- Tested and approved before use

### Retention and deletion

One rule per kind of data, from screenshots to exports, checked against the notice your employees acknowledged.

- One rule per kind of data
- Checked against the notice
- Effective date and reason

## For organisations whose data has to stay inside

Some teams cannot put project, time and people records in a shared cloud, however good the product. Akollo installs where their rules say it must, and keeps the evidence their auditors ask for.

- **Finance and insurance** — Change, IT and regulatory projects on your own servers, with an audit trail for internal audit and the regulator.
- **Public sector** — Projects and working time without an outside cloud, with retention that follows your records policy.
- **Healthcare** — Staff and project data inside your own network, with sensitive fields hidden or masked by role.

[Akollo for regulated industries](https://akollo.com/en/solutions/regulated-industries)

## What your security team can check for itself

No promises to take on trust. These come from the installation and the settings your administrators control.

- **Outbound connections** — Which external hosts connectors may reach, set once for the installation and narrowed again in each connector's own settings. (Security settings)
- **AI use by model** — Every AI request is recorded with the model it used and how much it used, never with the question or the answer. (Admin → Usage)
- **Privilege and configuration changes** — Who changed which role, permission or setting, and when, so a review starts from the record instead of from memory. (Admin → Audit records)
- **How long data is kept** — The retention rule for each kind of data, and how it compares with the notice your employees read and acknowledged. (Retention and deletion)

## An assistant that works inside your walls

Akollo AI can run on a model server you control, on your own provider key or on Akollo's models. It reads only what the person asking may see, and asks before it changes anything.

- “Summarise this week's status of the core systems programme.” — Five of seven projects are on track. The data migration has slipped a week and the network upgrade waits for a supplier. Shall I draft the steering note?
- “Who is free in November to help with the server upgrade?” — Three people with the right skills have room: Murat Öz from the second week, Lena Brandt and Selin Aydın from the third. Book them on the project?
- “Break the data centre move into tasks with owners.” — Here is a breakdown with 22 tasks in four phases, with owners suggested from the infrastructure team. Shall I add it to the project?

Data classified as restricted is answered only by a model inside your installation; if there is none, the request is refused. You can also switch the assistant off completely.

## Every Akollo module runs on your own servers

- [Projects & tasks](https://akollo.com/en/products/work) — Boards, timelines and portfolios.
- [Resource planning](https://akollo.com/en/products/people) — Capacity and skills for every team.
- [Time tracking](https://akollo.com/en/products/time) — Timesheets and approvals by unit.
- [Budgets & profitability](https://akollo.com/en/products/finance) — Project cost from the hours people log.
- [Reports & dashboards](https://akollo.com/en/products/insights) — Portfolio, hours and cost in one view.
- [AI assistant](https://akollo.com/en/products/ai) — On your own model, your key or switched off.

## Questions infrastructure and security teams ask

### Is the on-premise version a different product?

No. An installation in your data centre or your cloud account is the same product as our cloud, with the same modules and the same three languages. The difference is who runs the machines and who sets the network rules.

### Does Akollo need internet access to run?

No. An on-premise installation runs inside your network. Connectors cannot reach any external host until your administrators allow it, and the assistant can be limited to models inside the installation or switched off.

### Do we need graphics cards?

Not for Akollo itself. The application, background jobs and database run on ordinary machines. A graphics card only matters if you run the larger AI models on your own hardware, and the installation guide sizes that for you.

### Can we keep the AI completely inside our network?

Yes. Connect your own model server, directly or through the AI Connector, which connects out from your network so no inbound port is needed. Running every model, including the large ones, inside the installation is planned and already documented with hardware profiles.

### Can Akollo support see our data?

Only in a support session, which your organisation sees as a banner. A support session can read the pages it is allowed to, but cannot create, change, approve or delete anything, and sensitive fields stay hidden.

## Plan your installation with us

We will go through hosting options, sizing and network rules with your IT and security teams, and agree a first team to start with.

- [Book a demo](https://akollo.com/en/demo)
- [Read the docs](https://docs.akollo.com/en)


---

Source: https://akollo.com/en/solutions/on-premise · [Türkçe](https://akollo.com/tr/cozumler/kendi-sunucunuzda), [Deutsch](https://akollo.com/de/loesungen/eigene-server)