Agent Repo
Keep an agent’s instructions, saved memories and briefs in On My Radar, then explicitly load the context you need in a compatible tool.
An agent is more than a model. It’s the instructions you’ve refined and what it has learned about your work. The Agent Repo keeps both in On My Radar, so they no longer live in scattered files on one machine.
Check it out anywhere. Pick an agent and the folder you’re working in, and checkout adds a small pointer to that folder’s instruction file: CLAUDE.md for Claude Code, AGENTS.md for Codex, Copilot, Zed and others, or a Cursor rule. For Claude on the web, you paste the same text into a Project. The pointer tells your AI to load the agent from On My Radar when a session starts.
Bring saved context to your next session. Load the agent in a compatible tool, then read the saved memories and briefs relevant to the task. Saving and loading are explicit actions; a conversation is not saved automatically. Review the context before relying on it, because notes can be incomplete or outdated.
You stay in charge of its instructions. Every edit to the prompt is kept as a version, and by default a change an agent suggests waits for you to accept it. You can review what it remembers, pin what matters, and share an agent with someone as a viewer or an editor.
Find it at onmyradar.io/repo once you’re signed in.