Prime Agent

Prime Agent is a local terminal coding and research agent for work that continues across sessions. It runs commands and edits files in your project directory, while a persistent Python environment keeps working state available. Its hosted subscription and API-key routes send model requests to the selected provider; local execution doesn't imply offline inference.

Background sessions keep running after the terminal disconnects, and you can reattach later. Persistent goals, schedules and periodic wakeups let the agent continue longer tasks. Autonomous work can use turn, token and time limits, along with user-defined quality checks.

Built-in subagents can work in parallel or in the background. They exchange messages directly, return results to the parent agent and remain available for later tasks. This suits coding projects and research evaluations where separate pieces of work need coordination.

The agent can review a session and record focused lessons as memories, supplemental instructions or reusable subagent descriptions. Those refinements stay local to the session by default, leave the base system prompt unchanged and have snapshots for rollback. Recurring workflows can become executable Python skills for a project or personal use.

Custom provider support lets you specify model endpoints beyond the hosted login routes. The agent executes generated Python and project commands with your user permissions; its worker processes provide lifecycle isolation and recovery, not a security sandbox.