Connect Orvo to OpenAI Codex

Who this is for

Engineers who use Codex in the terminal or their editor and want to prep for meetings, capture decisions and plan follow-ups from the same place they work.

Before you start

• Codex CLI, the Codex IDE extension or the Codex app, signed in with a ChatGPT plan that includes Codex or with an OpenAI API key. See OpenAI’s Codex pricing page for which plans include the CLI and IDE extension.
• An Orvo account with web access (the free trial works).
• Optional: a personal token from Orvo Settings → Integrations, only if browser sign-in doesn’t work.

Your Orvo MCP server URL.

Fastest: paste a setup prompt into Codex

Codex can add Orvo to its own MCP configuration. Paste this prompt, then sign in with Orvo when your browser opens. No token needed.

1. Copy the prompt below. It’s also in Orvo under Settings → Integrations when you choose Codex.
2. Paste it into Codex’s chat and send it. Codex adds Orvo to its configuration. Approve its changes when it asks, and restart it if it tells you to.
3. When your browser opens Orvo, sign in, choose Read only (recommended) or Read and make changes, and click Authorise & Connect. Adding notes, actions or other changes needs Read and make changes.

Please connect Orvo, my personal CRM, to this app as a remote MCP server by adding it to your own MCP configuration.

- Server name: orvo
- URL: https://www.getorvo.com/api/mcp
- Transport: Streamable HTTP
- Authentication: OAuth sign-in through my browser. Orvo needs no API key or token, so do not add an Authorization header.

How to do it in Codex (the CLI, the IDE extension and the Codex app share ~/.codex/config.toml): run this command:
codex mcp add orvo --url https://www.getorvo.com/api/mcp
Codex detects that Orvo supports sign-in and opens my browser. If it doesn't, run "codex mcp login orvo", or tell me to run it in a terminal.
Then tell me to restart Codex and type /mcp to check that orvo is listed with its tools.

Please also:
- Replace any existing server called orvo instead of adding a second one, and keep it in my own user-level settings, not in a project file.
- When the sign-in starts, tell me what happens next: my browser opens an Orvo page where I sign in and choose "Read only (recommended)" or "Read and make changes", then click "Authorise & Connect". Adding notes, actions or other changes needs "Read and make changes".
- After I have signed in, check the connection and list the Orvo tools you can see. If they only appear after a restart, tell me.

If sign-in doesn’t work: a setup prompt with a token

Use this only where Codex can’t sign in through a browser, for example on a remote machine.

1. In Orvo, open Settings → Integrations, find AI assistants (MCP), choose Codex and open Use a token instead.
2. Click Generate token, then Copy setup prompt with token. The token is shown once, and generating a new one turns off the old one.
3. Paste the prompt into Codex’s chat and send it.

Please connect Orvo, my personal CRM, to this app as a remote MCP server by adding it to your own MCP configuration.

- Server name: orvo
- URL: https://www.getorvo.com/api/mcp
- Transport: Streamable HTTP
- Authentication: send the header "Authorization: Bearer YOUR_ORVO_TOKEN"

How to do it in Codex: the CLI, the IDE extension and the Codex app share ~/.codex/config.toml. Edit that file directly; don't use "codex mcp add", which starts a browser sign-in instead.
1. Preferred: store the token in an environment variable named ORVO_TOKEN that is set for my user whenever Codex starts (my shell profile on macOS or Linux, or "setx ORVO_TOKEN ..." on Windows), and add:
[mcp_servers.orvo]
url = "https://www.getorvo.com/api/mcp"
bearer_token_env_var = "ORVO_TOKEN"
2. Only if Codex can't see that variable (for example an app that doesn't read my shell profile), use a static header in ~/.codex/config.toml instead:
[mcp_servers.orvo]
url = "https://www.getorvo.com/api/mcp"
http_headers = { "Authorization" = "Bearer YOUR_ORVO_TOKEN" }
3. Run "codex mcp list" to check that orvo is configured. Then tell me to restart Codex and check its tools with /mcp.

Please also:
- Replace any existing server called orvo instead of adding a second one.
- Keep the token in my own user-level settings only, never in a project file, a repository or anything that is shared. Do not repeat the token in your replies.
- When you are done, check the connection and list the Orvo tools you can see. If they only appear after a restart, tell me.
- The token may be read-only. If Orvo refuses a change later, tell me to switch on "Allow changes with this token" in Orvo under Settings > Integrations (https://www.getorvo.com/setting?tab=integrations).

Set it up by hand: sign in with Orvo

1. Add the server from a terminal:
   Terminal Copy

   codex mcp add orvo --url https://www.getorvo.com/api/mcp

2. Codex detects that Orvo supports sign-in and opens your browser. If it doesn’t, run codex mcp login orvo.
3. Sign in to Orvo if asked, choose Read only (recommended) or Read and make changes, and click Authorise & Connect.
4. Start Codex and type /mcp to check that orvo is listed with its tools.

Set it up by hand with a token

1. In Orvo, open Settings → Integrations, find AI assistants (MCP), choose Codex and click Generate token. Copy it; it’s shown once.
2. Edit ~/.codex/config.toml directly (don’t use codex mcp add here: it would start the browser sign-in) and add:
   ~/.codex/config.toml Copy

   [mcp_servers.orvo]
   url = "https://www.getorvo.com/api/mcp"
   bearer_token_env_var = "ORVO_TOKEN"

3. Set the variable before starting Codex:
   Terminal (macOS / Linux) Copy

   export ORVO_TOKEN="YOUR_ORVO_TOKEN"

Read only, or read and make changes

With browser sign-in you choose on Orvo’s page. To switch, run codex mcp logout orvo, then codex mcp login orvo and pick the other option.

With a token, the Allow changes with this token switch in Orvo Settings → Integrations decides. Codex’s own approval settings still control whether it asks before running a tool.

Either way, deleting never erases anything: People and notes go to Recently deleted for 30 days, daily limits cap how much an assistant can change, and changes to a person (their profile, tags, notes and linked actions) show in their activity as made via AI assistant. How Orvo keeps you in control.

Things to ask Codex