Deploy from Codex CLI: connect Livops over MCP
Add a [mcp_servers.livops] table with url = "https://livops.uz/api/mcp" to ~/.codex/config.toml and run codex mcp login livops. Codex can then deploy, read build logs, explain a failure and roll back, and with default_tools_approval_mode = "writes" it asks before anything that changes.
By Livops team, the people who build and run LivopsUpdated
Codex CLI works in your terminal and your repository. Connected to Livops, it can also deploy what it wrote and read back why a build failed. This page has the config.toml entry for both ways of signing in, the approval setting that fits Livops’s tools, and what Codex can and cannot do with each.
What you need
- A Livops account, and a project or a GitHub repository to make one from.
- Codex CLI.
- For the token route only: an
lvp_API token from the team’s Settings → API tokens, created by the team owner.
Livops is a Streamable HTTP MCP server at https://livops.uz/api/mcp. Codex reads MCP servers from ~/.codex/config.toml, one [mcp_servers.<name>] table each, where a Streamable HTTP server needs a url[1].
Connect through the browser
[mcp_servers.livops]
url = "https://livops.uz/api/mcp"
default_tools_approval_mode = "writes"
Then:
codex mcp login livops
codex mcp login <server-name> signs in to a server that supports OAuth[1]. Livops opens in the browser: the app’s name, a team to pick, what the app will be able to do, and Allow. With two-factor on, your code comes first. Codex gets a key that lasts an hour, works only at /api/mcp and renews itself.
default_tools_approval_mode takes auto, prompt, writes or approve, and writes prompts for tools that are not marked read-only[1]. Livops marks every tool one way or the other, so with writes the reads run straight away and every change waits for you.
Connect with an API token
[mcp_servers.livops]
url = "https://livops.uz/api/mcp"
bearer_token_env_var = "LIVOPS_TOKEN"
default_tools_approval_mode = "writes"
bearer_token_env_var names an environment variable whose value Codex sends as the bearer token in Authorization[1]. Export the token before you start Codex:
export LIVOPS_TOKEN=lvp_xxx
The token never lands in config.toml, which matters if you use a project’s .codex/config.toml: Codex reads that file only for trusted projects[1], and it lives next to code you commit.
Browser sign-in or token: what can Codex do?
Codex works in one team, as you, with your role. A viewer only reads.
| Browser sign-in | lvp_ API token |
|
|---|---|---|
| Read projects, deploys, build and runtime logs, metrics | Yes | Yes |
| Deploy, roll back, restart, set variables | Yes | Yes |
| Delete anything | Yes, if your role allows | No |
| Read a database password or bucket key | Yes, if your role allows | No |
| Open a Payme or Click top-up link, redeem a promo code | Yes | No |
Card numbers, your password, two-factor, sessions, creating API tokens, deleting the account, handing a project over, project members and raw backup downloads have no tool at all.
The token column is the server’s rule, not the tool list’s: an lvp_ request that would delete, pay or read a password gets 403, through MCP or not.
A deploy loop in Codex
- Codex edits the code and pushes to the branch the project follows. Every push to that branch starts a deploy.
- “Follow the latest shop-api deploy until it ends”:
list_deploys, thenget_deploy_log, which returns up to 5000 lines. - If it fails: “Why?” Codex can read the log itself, or call
explain_failure, which asks Livops’s own assistant for the cause and a fix. The previous version keeps running meanwhile. - If the new version misbehaves after going live: “Roll shop-api back to the previous deploy” (
rollback).
To build again without a new push, deploy redeploys a project from its current source.
Safety notes
- Tool annotations are hints. The specification says a client must treat them as untrusted unless the server is trusted, and that a human should be able to deny a tool call[2].
writesis that human step for Livops. - Logs, commit messages and query results are passed to the model as data. A log line telling the model to do something is still only a log line, but Codex decides what it does next; keep the prompt for changes.
- A database connection URL and a bucket secret are real secrets. Over a browser connection Codex can fetch them, which is the reason to approve those calls by hand.
- Calls are rate limited per account.
Deploying from Codex on Livops
- Add
[mcp_servers.livops]with theurlanddefault_tools_approval_mode = "writes"to~/.codex/config.toml. - Run
codex mcp login livops, pick the team, press Allow. - Ask Codex: “List my Livops projects.”
- Push, and ask it to follow the build.
Also see MCP (AI clients) for every tool, and API tokens for what a token can do.
Questions
Where does Codex keep MCP servers?
In ~/.codex/config.toml by default, one [mcp_servers.<name>] table per server. A project can have its own .codex/config.toml, which Codex reads only for trusted projects.
How do I sign Codex in without a token?
Add the server with just its url, then run codex mcp login livops. Livops opens in the browser, you pick a team and press Allow. On a machine without a browser, codex mcp login has a --no-browser option.
Which approval mode should I use?
writes. It prompts for every tool that is not marked read-only, and Livops marks each tool: reads set readOnlyHint, everything that changes something sets destructiveHint. Reads then run without a prompt and every change waits for you.
Can I keep the token out of config.toml?
Yes. bearer_token_env_var names an environment variable, and Codex sends its value in the Authorization header. Put the lvp_ token in that variable, not in the file.
What can a token not do?
Delete anything, read a database password or a bucket key, pay, or change the variables a whole environment shares. The server refuses those for an lvp_ token whatever tool asks for them.
How do I disconnect Codex?
In the cabinet, Account, then Connected AI apps, then Disconnect. With a token, revoke it under Settings, API tokens.
Sources
- OpenAI — Model Context Protocol (Codex CLI) (learn.chatgpt.com). Checked .
- Model Context Protocol — Tools (specification 2025-11-25) (modelcontextprotocol.io). Checked .
Put your project online today
Connect GitHub or drop a folder. No card is asked for.
Connect CodexRelated
- MCP server for deploying apps: what to look forAn MCP deploy server lets Claude Code, Cursor, Codex or Copilot deploy, read logs and roll back on your hosting account. Judge one by how it signs in, how far a key reaches, whether its tools say which ones change things, and what you can see and revoke afterwards. Vercel, Render, Railway, Netlify and Livops all ship one.
- Deploy from Claude Code: connect Livops over MCPRun one command, claude mcp add --transport http livops https://livops.uz/api/mcp, then sign in through /mcp. From then on Claude Code can list your projects, deploy, read the build log, explain a failure and roll back, as you and only in the team you picked.
- Deploy from Cursor: connect Livops over MCPAdd Livops to ~/.cursor/mcp.json with just its address, https://livops.uz/api/mcp, and sign in through the browser. Cursor's agent can then deploy, read build logs, explain a failure and roll back, and it asks you before each tool call unless you change that.
- What Livops's AI assistant does when a build failsWhen a build fails on Livops, the previous version keeps running and one button asks the assistant why. It reads the failed build's log and your project's commands, says what went wrong and how to fix it, and may offer one fix as a button that changes nothing until you press it. It never deletes, never sees variable values, and does not edit your code.