A project
manager your
agent actually
operates.
Not a task manager your agent can read. A structured system it lives inside of — reading your backlog, dispatching work, and building memory across days and weeks.
Context-switching is where work dies. Qalatra keeps your tasks, notes, and agents in one place — so you return to work already done.
Four pieces.
One living system.
A real database for tasks.
SQLite on your machine. Projects, contexts, priorities, recurrence, due dates, energy levels, agent assignments. Your AI doesn't scan a markdown file — it queries a schema.
- → structured fields agents can reason about
- → full-text search across years of history
- → connect to anything with an MCP server
Daily notes with actual memory.
A structured journal stored in your database — not flat files. Daily stand-ups, meeting notes, half-baked ideas. Your agent reads and writes them directly, so yesterday's context is always in scope.
- → daily and weekly reviews that include everything
- → agent can append, annotate, or summarize
- → your notes live in the database alongside your tasks
Agents that operate, not just read.
Any folder with an agent.config becomes dispatchable. Assign one to a task and Qalatra runs it, captures the output, and surfaces it for your review. You come back to work already done.
- → works with any command-line agent — Claude Code, Codex, opencode
- → MCP server exposes hundreds of tools across connected systems
- → schedule, queue, or trigger on events
Habits, kept honestly.
Streaks, cadence, and your agent checking in. The tracker the app actually needs, not a dashboard for dashboards.
- → daily / weekly / custom cadence
- → agent nudges, never notifications
- → streaks and history your agent can reason about
Open a terminal.
Your context is already there.
Drop into Qalatra's integrated terminal for long-form planning, deep-dive conversations, or anything that needs more than a task card. Bring Claude Code, Codex, opencode — any MCP-compatible agent connects directly to your tasks, notes, and backlog. No setup. Your whole system is already in context.
Free and open source.
Everything included.
Integrated terminal
A real shell docked inside the app. Open fullscreen, run agents interactively, or leave it idle while you work tasks above it.
Local-first, yours
SQLite. Plain markdown notes. Everything on your machine, versionable, backup-able, portable. No telemetry.
Four views, one database
Priority, Project, Backlog, Habits. The same data under four angles.
MCP-native
A built-in Model Context Protocol server on :3457. Any MCP-aware agent gets structured access to your whole system.
Queue, schedule, dispatch
Agents run in the background on a schedule, triggered by a task, or queued manually. Jobs are tracked — you see what ran, when, and what came back.
Open source, MIT
The whole thing's on GitHub. Fork it, self-host it, write your own agents. The roadmap is public.
A synced web companion.
Your local brain, everywhere.
Qalatra stays local-first. But sometimes you're on a phone, at a conference, or on a borrowed laptop. The web app syncs your tasks and notes end-to-end-encrypted so your agent's context follows you.
still you → still e2ee → still your own backend if you want
Your agent is ready.
Give it somewhere to work.
v1.8.0 · public beta · time estimates, capacity planning, and event cleanup.