Codex / local MCP setup

GIVE CODEX THE RADAR.

Connect the open-source Radar MCP server once, then ask Codex to audit AI readiness, generate buyer questions, measure brand visibility, check supplied products, and compare saved runs.

Connect Radar →Read as Markdown

ONE COMMAND.

Run this from a terminal with Codex installed:

codex mcp add optifeed-radar -- \
  npx -y --package=optifeed-radar optifeed-mcp

Confirm the server is registered:

codex mcp list

In the Codex terminal UI, use /mcp to see connected servers. The Codex app, CLI, and IDE extension share the same MCP configuration on the same host.

FIRST PROMPT: FREE.

Run a free AI-readiness audit on mybrand.com with Optifeed Radar.
Explain every finding and do not start a paid visibility check.

The prompt maps to audit_store. It verifies the connection without an API key or AI-provider charge.

01 / QUESTIONS

Review the prompt pack

Generate the buyer questions mybrand.com should be visible for. Show them before running a visibility check.
02 / BRAND

Cap the spend

Check mybrand.com's AI visibility in quick mode. Use OpenAI only and cap total spend at $0.15.
03 / PRODUCTS

Name and order the products

Check these products for mybrand.com, best first: Product A, Product B. Cap total spend at $0.40.

Radar does not import a catalog. The supplied order is the merchant ranking it compares with observed AI recommendations.

PASS KEYS SAFELY.

For paid checks, put the provider keys you use in a local .env file in the project where Radar runs, or forward selected environment variables through Codex configuration. Do not place real keys in prompts or commit them.

[mcp_servers.optifeed-radar]
command = "npx"
args = ["-y", "--package=optifeed-radar", "optifeed-mcp"]
env_vars = ["OPENAI_API_KEY", "ANTHROPIC_API_KEY", "GOOGLE_API_KEY", "PERPLEXITY_API_KEY"]

Codex stores user-level configuration in ~/.codex/config.toml. Only include the variable names for providers you actually use.