Privacy

TODO(legal): this is an engineering description of the data flow, not a lawyer-reviewed privacy policy. It must be reviewed before launch. Everything it states about the system is accurate; what it lacks is the formal structure a privacy policy requires (controller identity, lawful basis, retention rights, transfer mechanisms, contact for data subject requests).

We store events, not code

Parsemend stores the error events your SDK sends: the exception type and value, the stack trace, breadcrumbs, tags, environment, release, and the contexts your SDK attaches. That data is yours, and it lives in a Postgres table partitioned by day.

Parsemend does not store your source code. The one exception is managed mode, described below.

Where your source code goes, per fix mode

Mode	Does Parsemend read your source?
Bring your own agent	Never. Your agent reads it where it already lives.
Local CLI	Never. The run happens on your machine.
Your API keys	Assembled into a context bundle and transited to your model provider. Not retained.
Managed	Read through a GitHub App installation, for the duration of the run.

If you cannot send source code to a third party, use bring-your-own-agent or the CLI. Those two never make a network call that contains your code.

Retention

Events are retained for 90 days by default and then dropped with their partition. Deleting a project deletes its events.

Model providers

In the three modes where you bring your own credentials, your prompts and context bundles go to the provider you chose — Anthropic, OpenAI, Google, or an Ollama endpoint you operate. Parsemend does not proxy, log or retain those payloads. Their handling is governed by that provider's terms, not ours.

Analytics

The site uses Plausible, proxied first-party. No cookies, no cross-site identifiers, no personal data. There is no analytics on the application itself.

Contact

Mail [email protected].