Agents / choose a surface

LET YOUR AGENT RUN THE RADAR.

Optifeed Radar can be operated by an AI agent through a local MCP server or a JSON-first command line. It runs on your machine, uses your provider keys, and preserves the evidence behind each result.

Connect with MCP →Use the CLI →

TWO WAYS IN.

01 / MCP

Natural-language operation

Connect the stdio MCP server, then ask your agent to audit a domain, generate buyer questions, run a visibility check, inspect products, or compare snapshots.

See all five MCP tools →
02 / CLI

Commands and structured JSON

Ask an agent with shell access to run Radar directly. The CLI supports free audits, paid visibility runs, product checks, reports, diffs, sources, and machine-readable output.

See CLI commands →

TESTED GUIDES.

01 / CODEX

Connect through Codex MCP

Add the published stdio server with one command, verify it with codex mcp list, and begin with the free readiness audit.

Use Radar with Codex →
02 / CLAUDE CODE

Connect in one project

Add Radar in Claude Code's local scope, inspect it through /mcp, then promote the setup only after the free audit works.

Use Radar with Claude Code →

START WITH FREE.

  1. Run the zero-key audit.

    Use audit yourbrand.com or ask the MCP server to call audit_store. It checks crawler access, llms.txt, structured data, metadata, and the sitemap without calling an AI model.

  2. Review the buyer questions.

    Generate or export the prompt pack before a paid run so the category and intent are visible.

  3. Set a spend cap.

    Visibility and shopping checks use your provider credit. Start with quick mode and an explicit max_cost.

  4. Inspect the evidence.

    Treat scores as point-in-time estimates. Open the captured answers, positions, cited sources, and absences underneath the headline number.