Sprintra gives your AI coding agent persistent memory, project structure, and decision traceability. Choose your setup path below.
Hosted SaaS. Sign up, connect your AI tool, and start in under 2 minutes. No installation required.
Best for teamsRun Sprintra locally with SQLite. Your data stays on your machine. One command to start.
Best for solo devsGo to app.sprintra.io and sign up with email or Google OAuth. You'll be guided through creating your organization and first project.
Install the Sprintra CLI globally. It's a lightweight API client (42 KB, 2 lightweight dependencies).
npm install -g @sprintra/cli
Open Claude Code and run these two commands. They install the Sprintra plugin via Claude Code's built-in marketplace system — same path you'd install any other Claude Code plugin.
# Inside a Claude Code session: /plugin marketplace add Sprintra-io/sprintra-mcp /plugin install sprintra@sprintra
Claude Code clones the plugin, registers its hooks, and enables the Sprintra MCP server. You get:
Authenticate once: paste a token from app.sprintra.io/agents when Claude Code prompts, then bind the current directory to your project:
# In your terminal (one-time): npx @sprintra/cli login # In your project repo (one-time per repo): npx @sprintra/cli link --project=<your-project-id>
Restart Claude Code (or run /clear) — your next session opens with full Sprintra project context already loaded.
Don't use Claude Code? Run npx @sprintra/cli connect to configure Cursor, Claude Desktop, or Windsurf with the Sprintra MCP server only (no SessionStart hook).
Ask Claude Code to interact with your project:
"Create a feature for user authentication""What should I work on next?""Record a decision: we're using PostgreSQL for the database""Show me the current sprint status"No sign-up needed. Scaffold a local Sprintra instance with the installer:
npx create-sprintra my-project cd my-project npm start
This starts the Sprintra server + dashboard at http://127.0.0.1:4000. Your data is stored locally in ~/.sprintra/data.db (SQLite). Open this directory in Claude Code — MCP auto-connects.
Sprintra registers as a local MCP server via stdio. Add it to ~/.claude.json:
{
"mcpServers": {
"sprintra": {
"command": "npx",
"args": ["@sprintra/cli", "mcp"]
}
}
}Claude Code will auto-discover all 17 consolidated tools on startup.
In your repo directory, ask Claude Code:
"Create a Sprintra project for this repo"This creates a project linked to your current directory. Sprintra auto-detects the project when you work in that directory.
Visit https://app.sprintra.io to see your project dashboard with 20+ views: Kanban board, sprint tracker, burndown charts, knowledge base, and more.
Install @sprintra/cli for direct terminal access to your project data — without going through an AI agent.
npm install -g @sprintra/cli # Authenticate sprintra login # Auto-configure your AI tool sprintra connect # Quick project overview sprintra context --project proj-abc123 # What should I work on next? sprintra next-work --project proj-abc123
The CLI auto-detects whether to connect to your local server or Sprintra Cloud. See the CLI Reference for all 20+ commands.
Sprintra @0.6.0 ships with sprintra transcript — a local SQLite + FTS5 index over Claude Code's existing JSONL files at ~/.claude/projects/. No byte duplication, no cloud upload. Your full conversation history becomes keyword-searchable in <100 ms.
# One-time backfill (indexes every Claude Code session you've had) sprintra transcript reindex # List recent transcripts sprintra transcript list --since 2026-04-01 # FTS5 keyword search across every session sprintra transcript search "drizzle parameter binding" # Disk usage / oldest / newest / retention policy sprintra transcript status
Default local retention: 90 days (configurable via SPRINTRA_LOCAL_RETENTION_DAYS). Raw transcripts never sync to the cloud — privacy by default. Cross-device handoff happens through 5 KB session digests instead. See Memory Layer docs for the full architecture.
Memory Layer
How user prompts, session digests, and local transcript search work together
Install Skills
8 AI workflow skills — /capture, /standup, /wrap, /decide, and more
CLI Reference
20+ commands for managing projects from the terminal
MCP Tools Reference
17 consolidated tools with method dispatch pattern
Tour the dashboard
20+ views for managing your project visually