kanman works inside your repositories, your tracker and your team chat. This page shows what protects that work, where your code and data go, and who else touches them. Everything here describes what runs today. When something is planned, we say so.
Your code and data stay where you choose
You pick where the work happens. The rules, the checks before a merge and the audit log are the same in every mode.
| kanman Cloud | Your own runners | Everything on your runners | Self-hosted edition | |
|---|---|---|---|---|
| Plans | All plans | Enterprise | Enterprise | Enterprise |
| Repository checkout and coding tool | A fresh machine per run in Frankfurt, Germany | Your network | Your network | Your network |
| Model calls of the coding tool | The model access your team chose | Your runner, with your model access if you want | Your runner, with your model access | Your runner, with your model access |
| kanman’s own model calls (drafting, review, answers) | kanman’s cloud, on the route your team chose | kanman’s cloud, on the route your team chose | Your runner | Your runner |
| Board, decisions, audit log | EU (Stockholm, Sweden) | EU (Stockholm, Sweden) | EU (Stockholm, Sweden) | Your servers |
Runner-only and self-hosted: with “Run everything on our runners”, kanman’s cloud never calls a model provider and never reads or writes your repositories. Pushes, merge requests, reviews and drafts run on your runners, with your model access and your code host tokens. The coding tool’s transcript and tool output stay on the runner; kanman only receives status, summaries, file names and links. When no runner is online, work waits. Nothing falls back to the cloud.
The self-hosted edition runs the whole application on your servers, with Docker Compose or on Kubernetes. Model calls go only through your runners, email goes through your own mail relay, and there is no public sign-up. Updates come from kanman’s registry or, with Self-hosted Plus, as an offline bundle.
Runners connect out over HTTPS with a workspace token. They need no inbound ports and no database credentials.
Credentials never reach the agent
The coding tool proposes changes; kanman’s host writes them.
- No tokens in the agent’s environment. The coding tool never holds your code host token, a model key or the value of a run secret. It reaches the model and your repository through a proxy on the run’s host, which adds the credentials on the way out.
- A credential scan before every push. Only the host pushes and opens the merge request, after scanning the diff for credentials. A hit stops the push.
- Short-lived access. Each run gets a token that works for that run only. With the kanman GitHub App, repository tokens are limited to one repository and expire after an hour.
- Your repository’s own configuration stays off. From the repository the agent loads instructions only: no project hooks, environment files or tool servers, unless your team allows them.
- Stored keys are encrypted. Your model keys and connection tokens are stored encrypted. API tokens are stored only as a hash and are never shown again.
Redaction before anything is stored
Text from runs, coding tools and models passes a redaction step before kanman stores it, and the exact secret values of a run are removed from everything the run reports. Your privacy rules (terms and patterns per workspace, team or connection) also mask personal and customer data before a model call, on kanman’s cloud and on your runners.
Text that nobody on your team wrote, such as issue comments, merge request descriptions, CI logs or dependency docs, reaches a model only as marked data with its source. It can never become an instruction for kanman. Text from people without write access to the repository or the team becomes, at most, a finding for a person to look at.
Limits live outside the model
A limit that exists only in a prompt is not a limit. What kanman may change, spend and merge is enforced by kanman’s host, by your code host’s protected branches and by the merge rule:
- kanman merges only when your CI is green and a separate reviewer session, which did not write the change, approved that exact commit. Your code host’s approval rules apply on top.
- Allowed repositories, denied paths and change size limits are checked while the coding tool works and again on the final diff.
- A run’s internet access is limited to what the work needs: the code host, package registries, kanman, the model and the hosts your team lists. Networks your team marks as production stay unreachable.
- Budgets per run, day and month stop work when they are used up. Stop all runs stops every run of a workspace within a minute, in the cloud and on your runners.
Access control and single sign-on
- Workspaces are sealed off from each other. Every table in kanman’s database checks workspace membership on each row, so a query only returns data of workspaces you belong to, even if a bug in the app asks for more.
- Roles. Owner, admin, member and viewer. Only owners and admins change model access, policies, members, tokens and runners.
- Single sign-on (Enterprise) with Microsoft Entra ID, Okta, Google or any SAML or OIDC provider. Enforcing it for everyone needs a passing sign-in test first. Sign-on events are in the audit log. Directory sync (SCIM) is coming.
- Removing a member ends their access immediately.
Audit log and export
Every policy change, answer, stop and resume, and every action kanman takes on your behalf, is an append-only entry with who did it and when. Entries cannot be edited or deleted by hand. Owners and admins export the log as CSV or JSON. Retention is a workspace setting from 90 days to 5 years; periods longer than a year are part of Enterprise. kanman also records each model call it makes (step, model, region, tokens, cost), without prompts or answers.
Data residency
- Application data (accounts, stories, runs, decisions, the audit log) is stored in the EU, in Stockholm, Sweden.
- Cloud runs execute on fresh machines in Frankfurt, Germany, which are discarded when the run ends. kanman keeps no copy of your repository.
- Model calls go to the provider your team’s route names. With EU-only processing switched on, kanman uses EU routes only, for example Amazon Bedrock in an EU region, and refuses every other route before a request leaves. Each run records the region its model ran in.
Deleted teams and workspaces are erased completely 30 days after deletion. Details: Retention and deletion.
Encryption
- All connections to kanman use TLS.
- Stored data is encrypted at rest by our hosting provider.
- Model keys, connection tokens and secret values are additionally encrypted before they are stored; API tokens are stored only as a hash.
Sub-processors
kanman uses a small set of providers. The binding list with data categories and transfer safeguards is on the sub-processor page, Annex 2 of the DPA.
| Provider | Purpose | Location |
|---|---|---|
| Supabase | Database, sign-in, file storage and server functions | Stockholm, Sweden |
| Fly.io | Cloud run machines (not used with your own runners) | Frankfurt, Germany |
| Cloudflare | Delivery of website, app and docs, DNS, attack protection, email forwarding | Global network |
| Resend | Sending emails | Ireland |
| Stripe | Payments and invoices | Ireland |
| Support mailbox | EU and USA | |
| Anthropic | AI assistance in operations and support | USA |
| Anthropic, OpenAI | Model calls, only when a team pays for model access through kanman | USA |
With your own key, your AI gateway or your runners’ model access, the model provider works under your contract and is not our sub-processor. We announce a new sub-processor to workspace owners at least 14 days before it starts. We do not use your content to train models, and we use model access only on terms that do not allow the provider to train on it.
Security documents
- Security whitepaper: architecture, data flows and controls in one document. Use Print or save as PDF for your files.
- Policies: access, change, incident, supplier and backup.
- Data processing agreement with the technical and organisational measures.
- Answers to standard security questionnaires are available on request: write to [email protected].
- The security pages of the docs explain each control with the settings behind it.
Status and uptime history
status.kanman.ai shows the live status of each component, its uptime over the last 90 days and the incidents of the past two weeks.
Responsible disclosure
Found a vulnerability? Write to [email protected]. Tell us what you found, how to reproduce it and what an attacker could do with it. German or English, both are fine.
A person reads every report, confirms it and keeps you posted until the fix is out. Please test only against your own workspace, leave other people’s data alone, and give us a fair chance to fix the issue before you publish. There is no paid bug bounty. Our machine-readable contact is at /.well-known/security.txt.