TaskNebula architecture
A conventional stack with agent-aware edges.
TaskNebula uses a precise, conventional stack: Next.js, PostgreSQL with pgvector, Redis, realtime collaboration, optional LiveKit, Docker Compose, localization, and a public roadmap.
Work graph on PostgresSSE realtimePublic roadmap
- Next.js
- PostgreSQL + pgvector
- Redis
- Realtime collaboration
- next-intl
- Docker Compose

- Deployment path
- Docker
- Locales
- 30
- AI posture
- Opt-in
- License
- MIT
01Stack map
A conventional stack, arranged for self-hosting.
TaskNebula places AI-agent governance over infrastructure many teams already operate: Next.js, PostgreSQL, Redis, optional LiveKit, and Docker Compose.
Backend
Data
Realtime
Ops
Trust
02System model
Realtime work graph, not a super-app.
The roadmap keeps the boundary clear: no chat, video, or wiki super-app; no full standalone calendar; and no native-mobile-first detour. The work graph remains the core.
01
Next.js standalone web
TaskNebula ships as a standalone Next.js web container with App Router, React, route handlers, and Docker-first runtime assumptions.
02
Postgres-centered work graph
The roadmap explicitly positions TaskNebula as a self-hostable Linear control plane that avoids sprawling infrastructure and centers the work graph around PostgreSQL.
03
Realtime without exotic sync
Issue create/update/delete events use SSE with Redis pub/sub and in-process fallback for single-instance installs. Yjs supports collaborative editing where wired.
04
Global product shell
The app ships 30 languages, browser/device auto-detection, native-name language switching, and RTL support for Arabic and Hebrew.
05
Trust and audit surfaces
RBAC, issue security levels, audit-log action types, signed webhooks, AGENTOWNERS, and approval gates are part of the public technical surface.
06
Roadmap is honest
The roadmap names shipped, partial, and open work, including tenant isolation, API-key auth, pagination, MCP, importer depth, and governance completion.
- 01
Browser / PWA
localized app
- 02
Next.js route handlers
Zod + auth
- 03
PostgreSQL work graph
issues + embeddings
- 04
Redis fan-out
SSE events
- 05
AI / integrations / agents
opt-in BYOK
03Roadmap posture
Shipped, partial, and open work are separated.
Technical evaluators can distinguish available capability from partial implementation and open roadmap work without inheriting an inflated maturity claim.
Foundation
Work-graph foundations
Labels, components, release records, resolution, issue rank, workflow enforcement, boards, structured queries, notifications, tenant isolation, API keys, pagination, and scale floor.
Next
Developer workflow and AI-native differentiation
Deeper code-host links, importer improvements, release notes, semantic search, Ask-AI completion, agent approval queues, MCP expansion, and coding-agent handoffs.
Longer term
Premium-gated OSS wedge
Custom hierarchy, field schemes, permission enforcement, portfolio plans, SLAs, configurable dashboards, proactive risk detection, and agent builder direction.
04Evaluation notes
Known constraints stay visible.
Documented gaps and staged work remain part of the evaluation; TaskNebula does not claim finished enterprise parity across every surface.
Built but still being wired
Ask/RAG, embeddings, MCP, agent engine, analytics, and importer work have visible roadmap follow-through.
Trust hardening track
Tenant isolation, API-key scopes, pagination, type-safety restoration, and governance completion are named as platform prerequisites.
Product direction
Self-hostable Linear, control plane for AI agents, Postgres-first operations, and no super-app sprawl.
Run it yourself
Pull the image, keep the data, wire your own AI keys.
TaskNebula is a self-hostable project-management control plane for teams that want inspectable operations instead of another black-box SaaS.