Ongoing development platform
AI development automation
An ongoing development platform exploring how coding tasks can be planned, delegated and verified across multiple AI providers, while keeping the developer in control.
How it fits together
- Request
- Development task
- Route
- n8n orchestration
- Project routing
- Execute
- Codex or Claude
- Persistent sessions
- Verify
- Validation
- Review and tests
The challenge
AI development tools are powerful individually, but each has different strengths, and using them well involves a lot of manual coordination. The aim is to explore how development work can be orchestrated end to end, with suitable tools doing each part and the developer staying in control. The working infrastructure is real, and the orchestration architecture is still being developed and refined.
What was built
- Remote execution of development tasks through REST APIs, on dedicated development worker infrastructure.
- n8n workflows that orchestrate tasks and route each one to the right project.
- Codex and Claude as active providers, with the provider and model chosen per task rather than tied to one vendor.
- Persistent AI sessions, and skills and instructions shared between providers so behaviour stays consistent.
- Planning, controlled implementation and validation steps, with review and testing stages being built out as the platform develops.
Technical highlights
- 01
Plan, implement, verify
Work is broken into planning, controlled implementation and validation, so changes are checked rather than accepted as generated. Review and testing roles are being refined as the architecture evolves.
- 02
Not tied to one vendor
Tasks can be run through different AI providers, allowing the provider and model to be selected for the job rather than tying the workflow to a single vendor. Local models are part of ongoing experiments.
- 03
Practical, not theoretical
The system is actively developed and tested against real software projects, providing first-hand experience of both the strengths and limitations of AI-assisted development.
What it demonstrates
Hands-on, current experience building with AI tools and agents Workflow automation with n8n and APIs Designing orchestration, validation and review processes A pragmatic view of where AI is, and isn’t, the right tool
Start with a conversation
Have something similar in mind?
Real-time data, unusual interfaces, integrations or automation: if it’s a problem software can solve, let’s talk about it.
No technical knowledge needed. No obligation.