- version
- 0.6.0
- pull requests
- 57
- automated checks per run
- 317
As of 10 November 2026
Product tour video
From requirement to release
Requirement
Written down, reviewed and versioned with a checksum.
Starter package
Rules and a stage plan for the coding agent.
Tasks
Each stage becomes a task on the board.
Result
The agent reports back with a pull request.
Approval
Bound to content and source revision.
Release
Deployed on click, only for the approved revision.
Problem
Coding agents are fast. Running several products alone still requires an overview: what is approved, at which revision, within which limits – and who decided last?
Solution
App-Leitstand isn't another agent platform. It's the governance layer a solo operator needs to let agents work safely.
- Requirements → reviewed versions with checksum → starter package for the coding agent → stages as tasks → results → approvals.
- Approvals bind to project, action, target, source revision and content; any change invalidates a decision.
- Project factory: generated starter packages (AGENTS.md, stage plan, licence, stack scaffolds) give the agent clear boundaries.
Making data visible
- Progress per project with a burn-up chart, pace per week and a linear forecast to the target date; stages from the starter packages as a bar.
- Honest, not prettified: every status change is logged by a database trigger; where no log exists yet, the line is dashed and labelled “derived”.
- Kanban board: checklists with progress on the card, due dates, time in column, WIP limit, keyboard shortcuts.
- Team progress from GitHub: progress per person from issues and pull requests, unassigned work, a chart with forecast – read only.
- Read-only guest access: team members see selected projects via Cloudflare Access; the server allows guests only read requests from a fixed allowlist – the one exception is a note to the owner, which arrives as a report. A test sends every write route as a guest and then compares all tables.
- All charts hand-drawn (SVG, no library, no external request): light and dark, German and English, explorable by keyboard, with a table for screen readers.
How I work: orchestrating, not just coding
I work as an orchestrator: I define requirements and boundaries, Claude Code implements them in stages, and every change comes back as a pull request. Nothing is merged until type checks, unit tests and a browser run across the whole interface are green; it is rolled out via GitHub Actions only after my decision.
That is how App-Leitstand itself is built – and how it steers my other projects.
Architecture and decisions
- MCP server (stdio and Streamable HTTP): Claude Code and Claude Desktop read tasks and approvals directly and report results back – one tool source for both transports.
- LangGraph pilot in Python with interrupts, SQLite checkpoints and a deterministic offline test model – labelled honestly, no provider call without a budget decision.
- Online only behind Cloudflare Access plus its own JWT check: two locks.
- Zero-cost architecture: self-hosted tools (Langfuse for tracing) instead of SaaS, Docker Compose instead of installers, ops scripts with dry run as the default.
Status
Version 0.6.0 (10 November 2026). App-Leitstand already steers my own development, including the neuefische capstone project, and has been running online behind Cloudflare Access since 9 November 2026. Setting it up at home takes a double-click: a starter checks Node.js, installs, builds and opens the browser; an illustrated guide and a handbook explain every menu item. An open-source release is in preparation.
Architecture
How my apps run
One small server, Cloudflare in front, GitHub for tests and delivery: the architecture behind the projects, with a diagram.
Download preview version 04.11.2026
Not fully tested yet – download with an access code only. Ask me for the code.