How Enderis handles your data
Last updated: November 7, 2026
This page says what Enderis stores, why it uses language models, and what you control. To learn how we protect your data, see Trust and security.
What we store
- Each workspace has its own database. It holds your agents, evals, playgrounds and runs. It also holds each run's model calls and the transcripts of the assistant and the authoring tools.
- Account and organization data is stored separately from workspace data.
- On the hosted service, our API and workers are in the United States, and we create new workspace databases there.
- An organization owner or admin can delete a workspace. This deletes its database, with its agents, evals, runs and playgrounds.
Why we use language models
Enderis uses language models for its own work, such as:
- Grading runs against the checks in your evals.
- Playing the simulated people in a test.
- Running the agents that Enderis hosts.
- Powering the assistant and the authoring tools, which help you write evals and playgrounds and improve your agent.
Each call sends the content for its task, such as a run's transcript, an eval's files or your request. By default, these calls go to third-party model providers, on our accounts.
We do not train models on your data without your permission.
A workspace admin can send most of these calls to your own model endpoint: grading, simulated people in text tests, the assistant and the authoring tools.
Self-hosted installs
With our default Helm chart, a self-hosted install keeps your data in your own database, model calls go to your own model gateway, and your data does not come to us.
Repository agents
A repository agent runs on your machine and calls its model provider with its own keys. Enderis does not make those calls, and your model keys are not sent to us.
Onboarding sends metadata such as file paths and variable names, not the contents of your files. A run sends its record to Enderis, including the agent's replies, its tool calls and their results. By default, the record also includes the prompts and replies of the agent's model calls. You can keep those on your machine by choosing a different capture policy there.
Production logs
Enderis reads production logs only if you connect a log source. You choose the capture policy and retention for each source. The default policy keeps the structure of the data, such as field names, types and sizes, and not message or tool content. Enderis removes secrets that it detects. On the hosted service, a policy that keeps values needs our approval.
Research sharing
Sharing for research is opt-in and off by default. Enderis copies a workspace's model calls and its feedback on insights to a separate research database only if a workspace admin turns sharing on. To remove copies, including after you turn sharing off or delete a workspace, email support@enderis.ai.
Questions
To ask a question, or to get the list of companies that process data for us, email support@enderis.ai. See also our privacy policy.