The default should be local
Most AI development tools require an API key before you can do anything. You sign up, enter payment details, configure environment variables, and then you can start working. That is backwards.
The default experience should be: clone the repo, run the command, start coding. No accounts, no API keys, no network dependency.
Why local models are good enough
Local models have gotten remarkably capable. For the tasks that matter most in day-to-day development — code completion, bug detection, refactoring suggestions — a 7B parameter model running on consumer hardware provides a usable experience.
Is it as good as the frontier cloud models? No. Is it good enough to be productive? Yes.
The gap between local and cloud models is closing faster than most people expect. What was frontier six months ago runs on a laptop today. Building for local-first means you automatically benefit from this trend.
The real advantage: latency
Cloud API calls have inherent latency. Even with fast connections, you are looking at 200-500ms round trips before the first token arrives. Local inference starts generating immediately.
For interactive use cases — autocomplete, inline suggestions, quick explanations — the latency difference is the difference between a tool that feels like part of your editor and a tool that interrupts your flow.
Zero-config as a feature
Every configuration step is a potential drop-off point. Every API key is a secret to manage. Every network dependency is a point of failure.
8gent ships with sensible defaults that work without configuration. You can connect cloud models for more capability, but you do not have to. The tool works out of the box.
This is not an ideological position. It is a product decision based on the observation that the best tool is the one people actually use, and people use tools that work immediately.
---