Security and data handling.

Emolysis is designed so that the work happens on your machine and production runs on your platform. Emolysis's own services hold your account, your plan, your usage and the dashboards you choose to publish. They are designed never to connect to Databricks, Snowflake, Microsoft Fabric or any other data platform of yours. Beyond that, they reach the provider you sign in with (GitHub, Microsoft or Google) for your account identity. The GitHub App is planned; once you install it, they also reach the repositories you choose.

That does not mean nothing leaves your machine. To write code, the agent sends prompts to the AI model provider you choose, and those prompts can include schemas and sample values. The list below says exactly what goes where.

When you connect a platform, you sign in with that platform's own login: browser OAuth for Databricks, Microsoft Entra ID for Fabric, and a key pair or programmatic access token for Snowflake.

The resulting credentials are kept in your operating system's keychain, its credential store. They are not sent to Emolysis. Emolysis acts with your account's permissions on the platform, no more.

We recommend connecting a development workspace, or a role with the narrowest permissions your pipelines need.

The Free plan is bring your own key, and every plan accepts your own key from the model provider you choose. The key is kept in your machine's keychain, and requests made with it go to that provider under its terms.

On paid plans you can use the plan's model usage instead of a key: included usage on Pro, Max and Team, usage credits past a limit, and metered usage on Enterprise. These requests pass through Emolysis's metering service, which is designed not to store or log the content of prompts or responses. It records token counts so usage can be billed against your plan.

What a provider does with a request is set by its own terms. Read them before you add a key. See pricing for how your own key and a paid plan's model usage fit the plans.

You sign in with your GitHub, Microsoft or Google account. The app signs you in through your browser in one click, with a short code as the fallback. The emo command line and the MCP server show a short code that you confirm in the browser. The code expires after 15 minutes.

Access tokens are short-lived. Refresh tokens are rotated each time they are used, and they are revocable.

A dashboard is the one thing you publish on purpose. When you publish, its charts and the data behind them are uploaded and stored by Emolysis. You choose who can open it.

A published dashboard is stored by our hosting provider: only what you publish, the rows its charts show, never your source tables. Each organization's dashboards are placed in Western Europe where our hosting provider honors that placement, without a residency guarantee. EU residency is planned for when our hosting provider supports it.

A dashboard keeps one stable link, and each publish adds a revision: a fixed snapshot of the rows its charts show. Charts published on their own get a snapshot link, which can have an expiry date. Revoking either deletes its published data. Who can open a link is fixed for its life: widening or narrowing it means revoking the link and publishing again under a new URL.

The GitHub App is planned. It asks for the fewest permissions its checks need, and you choose which repositories it can see.

If you prefer, a reusable GitHub Actions workflow, in preview, runs emo check in your own CI, with no app installed. It needs the emo command line, which is not released yet.

Signed desktop builds for macOS and Windows, with signed updates, are planned, so your machine can check that a build came from us. The Linux build is checked against its published checksum. No build is released yet.

The app keeps a session log of what the agent did: the requests, the files it changed, the tests it ran and what it deployed. The log is kept on your machine, and you can export it (planned). Organization-wide audit export comes with the Team plan (planned).

Emolysis has not been through a third-party security audit or penetration test, and it holds no compliance certifications. We plan to publish a fuller description of the security design before the first release.

If you think you have found a vulnerability, please tell us privately by email. Do not open a public issue or post details publicly.