Vibestrate is not a general-purpose agent framework. It is a local AI coding control layer that helps you supervise a crew of coding models without writing orchestration infrastructure from scratch.
Different jobs
| Question | Agent framework | Vibestrate |
|---|---|---|
| What are you building? | An AI application, service, or custom runtime. | A supervised AI coding workflow for your repo. |
| Who operates it? | Your application code and infrastructure. | The developer, from a local CLI, TUI, and Mission Control UI. |
| What is reusable? | Code abstractions, tools, memory, and agent runtime logic. | Flows: phase recipes with roles, gates, validators, and handoffs. |
| Where is the control point? | Usually in your custom code. | In the run: approve, reject, replay, and inspect each step. |
The practical difference
A framework asks you to build the orchestration layer. Vibestrate gives you a working layer for a specific problem: AI-assisted software work that needs supervision. It already has the concepts developers need for coding runs: task, flow, crew, worktree, review, verification, and trace.
That specificity matters. Vibestrate is not trying to be every possible agent system. It is trying to make AI coding work feel managed: the developer gives the direction, the crew does the work, and the flow keeps the process from disappearing into a single model.
The industry framing is moving the same way. As Google engineering leader Addy Osmani describes it, the developer's role is "evolving from implementer to manager... from coder to conductor and ultimately orchestrator." A framework leaves you building that orchestration layer yourself; Vibestrate is the orchestration layer for coding work.
Where Vibestrate is intentionally opinionated
- Local-first: the orchestrator runs on your machine.
- Human in the loop: risky moments pause for explicit approval.
- Model-flexible: each phase can use the model that fits it.
- Flow-centered: workflows are named, reusable, shared, and forkable.
- Auditable: tokens, cost, prompts, diffs, and decisions stay visible.
When a framework is the right answer
If you are building a customer-facing AI feature, a support agent, a voice agent, an autonomous browser worker, or a bespoke runtime, use a framework. If you are trying to keep your own coding work supervised across Claude Code, Codex, Gemini, local models, and shared workflows, use Vibestrate.
In short: frameworks help you build agents. Vibestrate helps you supervise AI coding work.
