Features / The two agents
Nodge runs two distinct kinds of agent, on purpose. One works with you while you code. The other works on its own, alongside your running app.
Every deployed app gets its own autonomous agent. It watches the app, responds to problems, and opens pull requests for a human to review. It is always present and only does work when you give it something to react to.
The agent can be triggered by a log error, a schedule, an incoming webhook, or a manual request. You decide what wakes it up.
When the agent opens a pull request, you review the diff on mobile, approve it, request changes, or chat with the agent about its reasoning. Approvals happen wherever you are.
The agent can pause and ask a person mid-task. You set how much autonomy each project gets, from full self-direction to a check-in at every step.
The autonomous agent runs in its own container, separate from your application. It cannot read your app's environment or secrets. It asks the platform for what it needs, and every request is governed.
Every token the agent spends is metered by the platform itself. A runaway loop cannot exceed the budget you set, because the limit is enforced on the platform, not trusted to the agent.
Two agents, two trust boundaries, one toolkit.
One is an interactive coding agent that works with you in the IDE. The other is an autonomous agent that runs alongside your deployed app. They have different jobs and different trust boundaries, so they are kept separate.
A log error, a schedule, an incoming webhook, or a manual request. You choose what wakes it up.
It opens a pull request for review rather than pushing changes straight to production. You approve, request changes, or discuss it first.
No. The autonomous agent runs in its own container, separate from your app, and cannot read its environment or secrets. It requests what it needs through the platform.
Every token it spends is metered by the platform, and your budget is enforced there. The agent cannot exceed the limit you set.
Yes. Autonomy is configurable per project, from full self-direction to a check-in at every decision.
