Feature request: fully integrated local multi-agent coding workflow
I’m a heavy power user doing deep planning plus direct repo/machine execution. The ideal product for me would be:
- multi-agent local-only execution
- planner and executor in the same IDE space
- full workspace and machine context
- persistent shared context between agents
- no fragile handoff between browser chat and IDE agent
- pricing/usage that supports heavy planning phases without request anxiety
Why this matters:
The current split between browser planning and IDE execution creates friction. The highest-value workflow is a single environment where I can discuss architecture, decide implementation, then have local agents inspect, edit, and run code directly on the machine while preserving shared context.
Current best approximation:
VS Code + agentic coding tools is close, but request limits on planning-heavy usage become a real constraint. I would strongly prefer an OpenAI-native solution that combines:
- high-quality planning/reasoning
- direct local execution in the repo
- persistent shared context in one place
- local-first operation rather than cloud-only execution
Concrete example:
I often do architecture/planning in browser chat, then separately hand off instructions to an IDE agent to implement locally. The missing piece is persistent shared context between planning and execution in one local-first workspace.
This would materially improve professional developer productivity.