I moved my MCP servers from Claude to Codex. Here's the cheat sheet

Full disclosure up front: I run PickAPlugin, a free directory of ChatGPT and Codex plugins, and I built a small converter while doing this move myself. You don't need it, though. Everything below works by hand. Same servers, different wrapper Claude Desktop and Cursor keep MCP servers in JSON. Codex keeps them in TOML, in ~/.codex/config.toml (or .codex/config.toml inside a trusted project). The servers themselves don't change at all. Only the config format does. Here's a server in Claude's JSON: {"mcpServers": { "context7": { "command": "npx", "args": ["-y", "@upstash/context7-mcp"] } }} And the same server in Codex's TOML: [mcp_servers.context7] command = "npx" args = ["-y", "@upstash/context7-mcp"] Each server becomes its own [mcp_servers.NAME] table. The CLI and the IDE extension both read the same file. Things that bit me env becomes an inline table. So "env": {"API_KEY": "..."} turns into env = { API_KEY = "..." }. Remote servers use url. Put any headers in http_headers. Keep tokens out of the file. For remote servers, bearer_token_env_var points Codex at an environment variable instead of pasting the token into config.toml. You don't have to touch the server. Local stdio servers run the same command everywhere. Or skip the file and use the CLI codex mcp add context7 -- npx -y @upstash/context7-mcp codex mcp list Where the old config lives macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json If you have a lot of them Converting one or two by hand is fine. If you've got ten, I put a paste-and-convert page here: MCP config converter. It runs in your browser and nothing gets uploaded. It goes the other way too, TOML back to JSON. The longer version of this guide is on pickaplugin.com.