A local-first control plane for running your AI dev team. Create tasks, assign them to AI agents, and stay in the loop with approvals — not manual prompting.
Built for developers using tools like Claude Code and Codex who want structure around their AI workflows: tasks, agents, runs, sessions, and visibility into what actually happened.
- Create a task from the dashboard
- The orchestrator assigns it to the right agent
- Manager agents can delegate subtasks to teammates
- Run activity streams back in real time
- You approve, revise, or retry — the agents execute
- Dashboard — React UI for tasks, agents, approvals, and live run activity
- Multi-agent orchestration — agents with roles, departments, and delegation scopes
- Task trees — manager agents break work into subtasks and delegate to specialists
- Agent drivers — pluggable support for Claude Code and Codex CLI
- Session persistence — agents resume context across runs on the same task
- Desktop app — Electron shell with git diff inspection per agent working directory
- CLI —
solostack run,solostack doctor, database backup, and more - Local SQLite — everything runs and persists on your machine
- Company memory — shared long-term context across agents and tasks
- Dependency scheduling — task graphs with ordering and blocking
- External inboxes — Telegram, Slack, and webhook-driven task creation
- Multi-tenant auth — team access and permissions beyond single-founder mode
npm install
npm run dev- Dashboard:
http://localhost:5173 - API:
http://localhost:3000
npm run desktop:devnpx solostack run # serve API + dashboard
npx solostack doctor # check local environment
npx solostack -h # all commandsapps/dashboard — React UI
packages/server — Hono API + SSE
packages/core — Orchestrator + agent drivers
packages/db — SQLite schema + repositories
packages/shared — Types + validation
packages/cli — CLI tool
MIT


