Skip to content
All MCP clients

PromptOT for Codex CLI

OpenAI's Codex CLI now works with PromptOT via the Model Context Protocol. Manage prompts, blocks, versions, and test cases without leaving Codex CLI.

Install snippet

Drop this into Codex CLI

Add the snippet to your Codex CLI config (TOML format). The tools are available immediately on the next `codex` command.

Config file:~/.codex/config.toml
[mcp_servers.promptot]
command = "npx"
args = ["-y", "@prompt-ot/mcp"]
env = { PROMPTOT_API_KEY = "pot_YOUR_KEY_HERE", PROMPTOT_MCP_CLIENT = "codex-cli" }

Replace pot_YOUR_KEY_HERE with the key you generate in your project's API keys page.

What you can do

A few of the 23 tools at your assistant's disposal

These are some of the most useful day-to-day. See the full catalog on the MCP overview page.

list_prompts

See every prompt in a project, with version + status.

create_prompt

Spin up a new prompt with starter blocks in one shot.

save_draft_version

Persist the current block layout as a new draft version.

publish_version

Promote a draft to production for the live API to serve.

diff_versions

Compare two versions block-by-block before publishing.

list_test_cases

Pull the test cases attached to a prompt for evaluation.

Quickstart

Three steps to PromptOT in Codex CLI

Step 1

Generate a key

Create an MCP-scoped API key from your project's API keys page.

Step 2

Paste the snippet

Drop the snippet above into ~/.codex/config.toml and restart Codex CLI.

Step 3

Ask your AI

Try “Show me my PromptOT prompts” in Codex CLI. The tools take it from there.

Ready to wire up Codex CLI?

Generate a scoped key, paste the snippet, and ask away. Or browse the other supported assistants.