Open source · MIT · Self-hosted
Self-hosted project management, with a real AI copilot.
TaskNebula feels like Linear, scales like Jira, and drafts work with you like a pair programmer. Bring your own OpenAI / Anthropic key — or run fully offline with the native planner.
$ curl -fsSL https://raw.githubusercontent.com/\
neuraparse/tasknebula/main/scripts/quickstart.sh \
| bash
→ pulling neuraparse/tasknebula:latest
→ spinning up Postgres 16 + Redis 7 + LiveKit
→ generating AUTH_SECRET (openssl rand)
→ running migrations
→ http://localhost:3000
✓ first-run wizard ready
✓ admin account creation
✓ AI is off by default — enable per workspace3 min
Self-host setup
30+
Permission types
63+
Audit-log actions
MIT
Open licence
Capabilities
Everything a high-performing team needs.
Six product surfaces — each opinionated, each shippable, each maintained as a first-class concern.
Project management
Kanban with drag-and-drop, stories / tasks / bugs / epics / subtasks, sprints, burndown, velocity, custom fields, issue links, attachments, custom workflows and transition rules — backlog grooming and roadmap views included.
Collaboration
Real-time presence and a live activity feed. @mentions, watchers, reactions, threaded comments, email digests. LiveKit voice rooms for ad-hoc calls, project-scoped chat, and an in-app notification bell with deep-links straight to the work.
Admin & governance
30+ granular permission types and 63+ audit-log action types. Role-based access with issue security levels. API keys, signed webhooks, OAuth (GitHub, Google, custom). Multi-org support with per-org plans and feature flags.
AI assistant (opt-in)
AI is off by default. Enable per workspace from Settings → AI & Agents in under 30 seconds. Draft-with-AI auto-decides single ticket vs checklist; per-issue assist summarises, rewrites, suggests, or proposes labels. Native heuristic planner ships built-in.
Analytics
Burndown, velocity, cycle time, throughput, lead-time distributions. Project health scorecard, CSV / JSON export, sprint retrospectives view, and cross-project rollups when you need to see the whole programme at once.
Developer experience
⌘K command palette, keyboard shortcuts everywhere, dark mode, mobile responsive layouts, route-level skeletons (no blank loads), and a one-command production deploy via Docker Compose.
AI & Agents
Off by default. On in 30 seconds.
Bring your own platform-level OpenAI / Anthropic key from Admin → Agent control, or override per workspace from Settings → AI & Agents → Quick setup. Keys are AES-256-GCM encrypted with redacted previews and audit-logged rotations.
- Draft-with-AI — one prompt becomes a single ticket or a structured checklist. Pick what you want before bulk-creating.
- Per-issue assist — summarise with comments, rewrite the description, suggest next steps, or propose labels. One click to apply.
- Model profiles — save provider + model + temperature + max tokens + reasoning effort with full revision history.
- Fails gracefully — bad key, rate limit, model unavailable becomes an in-app notification with a deep-link to the right settings page.
Supported providers
gpt-4o · gpt-4o-mini · gpt-4-turbo · gpt-4 · gpt-3.5-turbo · o1 · o1-mini
claude-opus-4-7 · claude-sonnet-4-6 · claude-haiku-4-5 · claude-3-5-sonnet · claude-3-5-haiku · claude-3-opus
Built-in heuristic planner — no API calls, deterministic, ships in every install
Architecture