Codex and X

Prepare X posts with Codex.

A person confirms before anything posts by default. A workspace owner can explicitly enable auto-publish for this agent in this workspace.

Check channel access first

No platform review needed: connect your X account and post. X charges per post, so paid plans include a monthly X allowance; Free doesn't include X.

Read all X formats and limits

Connect your agent

OAuth:

codex mcp add heyeveryone --url https://heyeveryone.ai/mcp
codex mcp login heyeveryone

Personal access token:

export HEYEVERYONE_TOKEN=hey_pat_...
codex mcp add heyeveryone --url https://heyeveryone.ai/mcp \
  --bearer-token-env-var HEYEVERYONE_TOKEN

or in ~/.codex/config.toml (or a project's .codex/config.toml):

[mcp_servers.heyeveryone]
url = "https://heyeveryone.ai/mcp"
bearer_token_env_var = "HEYEVERYONE_TOKEN"

In the TUI, /mcp lists connected servers.
IDE extension: gear menu > MCP servers > Add server > Streamable HTTP.
The old experimental_use_rmcp_client flag is no longer needed for HTTP servers.
Codex speaks the initialize handshake (MCP 2025-11-25) unless its 2026-07-28 feature flag is on. heyeveryone serves both.

Sources: learn.chatgpt.com: MCP in Codex, Codex config reference.

Draft, review, then confirm

Connect your channel in heyeveryone and grant the agent access to its workspace.
Ask: “Draft a X post from these notes. Check the channel rules and show me the content link. Do not schedule or publish.”
Open the draft and inspect it. Ask the agent to request a review; a person creates the review link or reviews it personally.
Review the current text and approve it if sign-off is required. Confirm the channels and time before publishing.

Access and credits

Agent access is included in every plan. MCP and API calls do not cost credits; actions that run heyeveryone’s own AI do, with cost shown first. Your agent provider may charge separately.

Full Codex guide