From local development to verified production recovery—without switching tools.
StackCendra is an AI-native engineering workspace designed to understand source code, configuration, local environments, remote infrastructure, deployments, and production telemetry. Its first goal is deliberately focused:
Detect, reproduce, and resolve environment-related failures across local development and production.
Live: stackcendra.com
| Debug session with AI assistant | Project Git + CI/CD status |
|---|---|
![]() |
![]() |
Screens show sample workspace data.
StackCendra is in Phase 0: Foundation and product architecture.
The web app runs on Next.js App Router and is deployed at stackcendra.com. What is real today:
- Sign-in with GitHub or Google. Email/password is still concept UI.
- GitHub and GitLab integrations: access tokens are encrypted (AES-256-GCM) before storage and only decrypted server-side; GitLab tokens auto-refresh. The integrations page shows live CI/CD activity from connected repositories.
The dashboard, project views, debug session and AI assistant are still interface concepts running on sample data. They are not yet connected to a desktop agent, container runtime, cloud account, or AI investigation service. See Concept UI Screens for the running screen inventory.
The complete Phase 0–13 planning documentation is versioned in docs/wiki, which remains the canonical review history. The same reviewed page set is published to the hosted StackCendra Wiki.
Start with:
- Product vision and scope
- Phase 0 plan
- Phase 0 execution backlog
- Phase delivery framework
- Complete capability roadmap
- Wiki review guide
- System architecture
- Security and trust model
- Toolchain setup record
- Release 0.1: project discovery
- Architecture decisions
Discover project
→ understand configuration
→ generate and validate an environment
→ develop and test
→ review and deploy through controlled flows
→ observe runtime behavior
→ diagnose incidents with evidence
→ reproduce failures locally
→ verify recovery and prevent recurrence
| Technology | Responsibility |
|---|---|
| Next.js and TypeScript | Web experience and shared product interface |
| Tauri 2 and Rust | Desktop shell, local discovery, terminals, credentials, SSH, and Docker access |
| Go | Infrastructure control plane and provider integrations |
| Python | Evidence-driven repository and incident intelligence |
| NestJS | Presence, collaboration, WebSockets, and notifications |
| PostgreSQL | Core product state and resource relationships |
| Temporal | Durable, approval-based automation |
| OpenTelemetry | Correlation across code, deployments, configuration, and runtime behavior |
These are target responsibility boundaries, not permission to deploy every component as a separate service immediately. StackCendra begins as a modular system with the smallest useful number of runtime units.
- Build one complete workflow, not twenty shallow products.
- Every diagnosis must link claims to evidence.
- AI proposes; policy validates; humans approve; constrained runners execute.
- Secrets do not enter ordinary prompts, logs, Git, or browser storage.
- Every release must be independently demonstrable and useful.
- Local and customer-hosted execution handle sensitive or expensive work.
- Architecture evolves only when measured requirements justify additional complexity.
| Release | Promise |
|---|---|
| 0.1 | Detect and explain local projects |
| 0.2 | Understand configuration requirements |
| 0.3 | Compare environments and prevent drift |
| 0.4 | Securely connect to remote hosts |
| 0.5 | Automate repeatable remote actions |
| 0.6 | Build health-gated deployment flows |
| 0.7 | Correlate Git, deployment, and configuration |
| 0.8 | Inspect Docker and Kubernetes |
| 0.9 | Diagnose incidents using evidence |
| 1.0 | Reproduce production failures locally |
Implementation instructions will be added after Phase 0 establishes the monorepo and reproducible local development environment.

