For Claude Code users

Stop Claude Code from
forgetting your project.

Sprintra is the MCP server that gives Claude Code persistent memory of your sprints, decisions, and knowledge base — auto-loaded into every session. No more re-explaining yesterday's work.

$npx create-sprintra
60-second install · MIT licensed · Free local mode
Featured on· MCP Registry· Cursor Directory· Glama· Smithery
What Claude Code sees on your next session
# Sprintra Memory — Session Briefing

## Active Sprint
Sprint 28 · "Onboarding polish" · 5/8 stories done · ends Friday

## Recent decisions (last 14 days)
• Switched from PostHog free → Pro (need Group Analytics)
• Reject Performance Max ads — too expensive at our budget tier
• Use medium.com/p/import for cross-posts (auto-canonical)

## Pending asks from last session
• Demo video Fiverr brief (started, not posted)
• Twitter creator outreach (3 candidates identified)

## Open questions
• Should we self-host PostHog reverse proxy for ad-blocker bypass?

Auto-generated from your activity. No prompts to write, no context to paste — Claude Code just knows where you left off.

Three things Claude Code can do with Sprintra wired up.

Remember every decision

Architecture decision records (ADRs) with automatic conflict detection. Ask Claude what you decided about JWT 6 weeks ago — it actually answers.

Track sprint state

Active sprint, current stories, blockers — Claude Code reads them and writes back as work progresses. No more 'what was I doing?' on Monday morning.

Cross-IDE memory

The same project workspace is readable from Cursor, Codex, Antigravity, and Gemini CLI. Switch IDEs mid-feature without losing context.

How a Claude Code session looks with Sprintra.

  1. 1

    You start a Claude Code session.

    Hooks fire automatically. Sprintra loads a structured briefing into Claude's context — recent activity, active sprint, recent decisions, pending questions.

  2. 2

    You work. Claude writes to Sprintra as it goes.

    When you say 'log this decision' or 'create a story for the auth bug,' Claude calls Sprintra MCP tools directly. State persists immediately.

  3. 3

    Session ends. Sprintra captures a structured digest.

    Stop hook fires. Sprintra writes a digest of what happened, what was decided, what's pending — at zero LLM cost. Available on the next session.

  4. 4

    Next session. Pick up where you left off.

    New briefing reflects every change since last session. No 'what was I working on?' No re-explaining context. Just continue.

Install in 60 seconds.

One npx command. Auto-registers with Claude Code. Local-first SQLite — your data stays on your machine.

$npx create-sprintra

Free Solo Pilot · MIT licensed · No credit card · No telemetry

FAQ

How does Sprintra give Claude Code persistent memory?

Sprintra runs as an MCP (Model Context Protocol) server that Claude Code connects to on every session. It captures every prompt, decision, and session digest at zero LLM cost using Claude Code's built-in hooks (UserPromptSubmit, Stop, PostToolUse). On the next session, Claude Code automatically receives a briefing with recent activity, active sprints, and recent decisions before you type anything.

How do I install Sprintra for Claude Code?

Run `npx create-sprintra` in your project directory. The installer auto-registers Sprintra MCP with Claude Code, generates a CLAUDE.md from your existing git history, and seeds an initial project workspace. Total time: about 60 seconds. Local-first SQLite, no account required.

Is Sprintra free? What's the catch?

Solo Pilot tier is free forever — local SQLite, all 20 MCP tools, MIT licensed. No credit card. Hosted Team tier is $5 per seat per month for collaborative teams who want sync across machines plus dashboard access. The MCP server is open source on GitHub.

Does Sprintra work with other AI coding agents besides Claude Code?

Yes — Sprintra MCP works with Claude Code, Cursor, Codex, Antigravity, and Gemini CLI. The same project workspace is readable and writable from any of these agents through MCP. Move between IDEs without losing project context.

What's the difference between Sprintra and claude-mem?

claude-mem is a memory-only plugin that compresses tool-call observations into a vector store. Sprintra is a full project workspace — memory plus sprints, decisions, dependency graphs, and a knowledge base. Memory is the floor; project context is the goal. See sprintra.io/vs/claude-mem for an honest side-by-side.

Will Sprintra capture sensitive code or credentials?

No. Sprintra captures structured project data (prompts, decisions, sprint state) — not raw code or environment variables. In local mode (Solo Pilot), nothing leaves your machine. In hosted mode, per-user privacy is enforced by row-level security; cross-user reads require admin role. Three opt-out flags let you disable any capture layer.