Add the server
Codex reads MCP servers from ~/.codex/config.toml. A project file, .codex/config.toml, is used when Codex trusts that project. The ChatGPT desktop app, the Codex CLI and the IDE extension share the file.
Add Spekkr from a terminal:
codex mcp add spekkr --url https://spekkr.com/mcp
Or add this table to the config file. An entry with a url is a Streamable HTTP server.
[mcp_servers.spekkr]
url = "https://spekkr.com/mcp"
Spekkr does not require a token to search, read or pull. When the entry has an address and no credential, Codex can connect without authentication.
In the desktop app, open Settings, then MCP servers, then Add server. Choose Streamable HTTP, paste https://spekkr.com/mcp, save, and restart. In the IDE extension, open the gear menu, then MCP servers, and do the same, then restart the extension. In the Codex terminal, /mcp lists the servers that are connected.
Checked 20 November 2026. Codex's MCP docs.
Without connecting
An agent that can open a web address does not need the server. Paste one line. The address is that spec's start note.
Build the spec at https://spekkr.com/juno/waitlist/start.md.
juno/waitlist is a waitlist spec. The note tells the agent to fetch the files, record the choices and build.
Search, then pull
With the server connected, paste this:
Search Spekkr for a waitlist spec, then pull juno/waitlist with database=sqlite.
Spekkr tells the agent to search with search_specs, read get_spec, then call pull_spec with the person's choices. Write the files into a folder and follow BUILD.md.
juno/waitlist lists database as postgres or sqlite. The default is postgres, so the prompt sets sqlite. A choice the spec does not offer is refused. Anything left out keeps the spec's default.
Search, get and pull need no sign-in. Publishing, unpublishing and whoami do: the agent opens a Spekkr page, and Spekkr emails a code. There is no password. Questions: support@spekkr.com.
The other tools, and the limits, are on the docs page. Every agent is listed on the setup index.