Privacy Policy

Last updated: June 5, 2026

1. The Sprintra connector for Claude

When you connect Sprintra to Claude (through the Claude connector directory or as a custom connector), the connector reads and writes the project data in your Sprintra workspace — features, stories, sprints, decisions, documents, and comments — so Claude can do the project-management work you ask it to do.

The connector does not collect your conversations or chat history, does not read the files on your computer, and does not access Claude's memory. It only performs the project actions you direct, scoped to your own organization.

Access is authenticated with OAuth 2.1. Tool responses are size-limited (and may be truncated with a note to fetch more) so no large or unexpected data is returned, and the connector stores no conversation content of its own.

2. What we collect

Self-hosted Sprintra (local SQLite). Zero. Nothing leaves your machine. The CLI doesn't phone home and there is no product telemetry.

Sprintra Cloud (api.sprintra.io). Only what you explicitly hand us:

  • Account record (email, hashed password, or your OAuth identity from Google or GitHub) via Better Auth.
  • Project data you or your AI agent create — features, stories, decisions, documents, notes — the things Sprintra exists to store.
  • Session digests your AI agent writes when a session ends (see §3).
  • Basic usage analytics from our website and dashboard (see §6).

3. How the Memory Layer works

The Memory Layer is a local Claude Code plugin — separate from the connector above. It captures your prompts and end-of-session summaries so your next session has continuity. These are scoped per-user — your teammates can't read them. Cloud digests rotate after 30 days unless you pin them; raw Claude Code transcripts stay on your machine and never sync (local index, ~90 days). You can disable any capture with three flags in ~/.sprintra/config.json (capture_user_prompts, capture_session_digests, capture_transcript_index), a strict_local_memory mode that stops all cloud egress, or a <private>…</private> tag on any prompt to skip it entirely.

4. What we don't do

  • We don't sell your data.
  • We don't run ads.
  • We don't train models on your project data.
  • The connector never reads your conversations, chat history, or files.

5. Storage & sub-processors

Cloud data is stored in PostgreSQL on Supabase (US region) and served from Fly.io; the website runs on Vercel. Self-hosted data lives on your filesystem under ~/.sprintra/.

Semantic search and recall send the relevant text to Google's Gemini embeddings API to compute vectors. In strict_local_memory mode, Memory Layer text is never sent off your machine.

6. Analytics & cookies

Our marketing website and dashboard use PostHog (product analytics) and Google Analytics to understand usage and improve the product. These set cookies and collect standard usage data such as pages viewed and actions taken. We do not use this data for advertising and we don't sell it. The Sprintra connector and the MCP API themselves do not use analytics cookies.

7. Your rights & contact

You can request access to, export of, or deletion of your account and project data at any time. Privacy questions: privacy@sprintra.io. Product or connector support: support@sprintra.io.