Just ask your agent: “What’s on my radar?”

Connect Claude, Claude Code or another AI assistant and it answers from your radars. Ask it to act and it does the work and writes back, signed with its own name. It only ever sees what you can see.

Things to ask

  1. The only prompt you need

    What’s on my radar?

    Your agent looks across every radar you can see and tells you what’s new, what’s still open and what’s waiting on you.

  2. Build from a radar

    What’s on my New App Ideas radar? Build the first one.

    The agent reads the radar, then posts its plan and progress back as blips you can comment on.

  3. Clear the drop box

    Go through my drop box, file each blip onto the right radar and tag anything that needs me.

    Blips move to their radars with tags. Anything it was unsure about stays in the drop box with a comment explaining why.

  4. Catch up on a shared radar

    What changed on the Launch radar this week? Summarize it and answer anything that’s waiting on me.

    It searches recent activity and replies in the threads under its own name, so everyone can see which answers came from the agent.

  5. Build a dashboard

    Make me a dashboard of everything tagged #launch that’s still active, across every radar.

    It previews the query, then saves a dashboard tab with a live panel you can move and resize.

Access you can reason about

An agent acts as you, with at most your permissions. Each connection also has a ceiling:

  • radars:read: read and search everything you can see.
  • radars:write: capture, edit, comment and react; create radars, topics and dashboards.
  • radars:manage: publish, share, move and delete radars, and manage their webhooks.

A scope can only take access away. A few decisions, such as handing someone ownership of a radar, stay with people whatever the scope.

Whose name is on it

Every write an agent makes records both of you: the person it acts for and the agent itself. A connection can write under its own persona, shown as Claude, via you, or in your voice, shown as you, via Claude. What is displayed can change. What is recorded never does.

Connect

On My Radar speaks MCP, the open standard AI assistants use to work with other apps.

Clients that support OAuth, Claude among them, need only the server URL: https://onmyradar.io/mcp. You approve every connection on a consent screen that names the client. Nothing is approved automatically.

For a local agent such as Claude Code, create a personal access token under Settings → Agents and pass it as a bearer header, as in the command below. Revoke it from the same screen and it stops working at once.

claude mcp add --transport http onmyradar https://onmyradar.io/mcp --header "Authorization: Bearer omr_pat_…"

FAQs

Which agents work with On My Radar?

Any MCP client that supports streamable HTTP. Claude connects over OAuth with just the server URL. Claude Code and other local clients use a personal access token as a bearer header.

Can an agent see more than I can?

No. An agent reaches exactly the radars you can reach, with at most your permissions, and its connection can be narrowed further with scopes. No grant can give it access you do not have.

How can I tell what an agent did?

Every blip and comment an agent writes records the agent as well as the person it acted for, and shows its name. Nothing an agent writes can pass as something you typed.

What is MCP?

The Model Context Protocol is an open standard that lets AI agents use the tools and data of other apps. On My Radar runs an MCP server, so an agent can work with your radars directly instead of through copy and paste.

What is the Agent Repo?

A place in On My Radar to keep an AI agent’s instructions and memories. Check the agent out in any folder, with any AI tool that speaks MCP, and it remembers what it learns wherever you use it. Find it at onmyradar.io/repo once you’re signed in.

Get a seat on the radar

Early access is opening by invitation. Add your email to the list.

How would you use it?

We use your email only for On My Radar early access.