ScalePlane · Own product
My interpretation of a software factory: workflows structure the work, triggers start it and agents handle tasks. I set the direction, review results and develop the platform further.
Software factory
Built as a solo developer with AI agents. ScalePlane connects the individual steps into a system I use to develop the product itself.
A task or trigger starts the work.
A workflow provides tasks and context.
Agents work with code and tools.
Tests and results make changes verifiable.
I review the changes and decide on the next step.
Currently in pre-release and used in my daily development work.
Project in numbers
Stand 20.09.2026: ein gemeinsamer Snapshot von Code und Entwicklungshistorie.
Applications and libraries
Tests and associated fixtures
Counted separately
Integrated changes
Excluding merge commits
The count covers ScalePlane’s code. Source code includes human and AI-assisted development. Tests, test data and generated files are reported separately. Blank lines, comments, dependencies, third-party code and build output are excluded from source code.
A further 51,825 lines of tooling and infrastructure code are counted separately and excluded from the source-code figure. The count also includes older product components that remain in the codebase.
Development history is counted without duplication. Merge commits, migration-only changes and changes outside the product are excluded from the respective totals. Pull requests and commits are different measures.
Counted with scc 4.1.0 on 2026-09-20. Source revision: 6565e7737057. The figures show scope and development history, not productivity or quality.
Brownfield, not a blank project
ScalePlane grows in an environment where models and agent tools change constantly. Existing features still need to work together: interface, API, database, agent runtimes, workflows and integrations.
That is brownfield development: changing a system with existing rules, data and dependencies. A new feature often touches several parts at once — from the data model to displaying a running task.
I see first-hand how well agents handle large codebases, what context they lack and which checks make changes reliable.
Engineering
A model alone does not deliver a project. It needs context, tools and a reliable workflow. In ScalePlane, this environment — the harness — connects different agent runtimes with tasks, repositories and traceable results.
Agents need an environment where they can actually work. That includes provisioning and cleaning up sandboxes, assigned access and handling secrets. The code implements separate credentials and masking of known secrets in output, among other measures.
Long tasks involve dropped connections, restarts and new instructions. Persisted state, recovery and verifiable results are therefore dedicated parts of the system. This is where the realities of working with agents become apparent.
Technology
The product also includes deployment, data storage, access and monitoring of running systems.
API, business logic, and persisted task and execution state.
Web interface, tools and the runtime connecting agents to the platform.
Servers, containers and configuration for running the platform.
Azure Key Vault for infrastructure secrets, Azure Blob for Terraform state and Terraform for SigNoz dashboards.
Working with you
I test what I build here myself — from the first agent task to running the platform. I bring that experience to your projects.
Together, we can clarify which tasks agents should handle, which tools and access they need, and how your team reviews the results. From the first experiment to integration with your existing IT.