Connect WarmySender to ChatGPT and Codex

WS
WarmySender Team
Updated Sep 6, 2026

On ChatGPT, adding your own connector lives behind developer mode, and the write actions a campaign needs — creating, launching, pausing — are a ChatGPT Business, Enterprise or Edu capability. A personal Pro account can connect a custom server for reading only. So if you are one person on ChatGPT Plus, Codex is the path that actually works — and Codex is already included with your paid ChatGPT plan. Setting it up takes two lines.

Both routes end in the same place: your agent talks to your own WarmySender workspace at https://warmysender.com/mcp and can build, launch and manage your cold email, LinkedIn and Instagram campaigns in plain language. And on both routes the safety line is identical — your agent never sends a message itself and can never raise a limit. It writes the campaign and hands it to WarmySender's scheduler, which paces every send inside safe daily caps and the gradual ramp no matter who pressed go.

Which path do I need?

What each ChatGPT plan can do with your own MCP server, and the route we recommend.
Your ChatGPT plan Add your own connector in ChatGPT on the web? Can it create and launch campaigns? Route we recommend
Free No No Codex is not included either — use Claude or another agent, or upgrade
Plus Developer mode is listed for Plus, but full connector support is not — treat the web route as unavailable Not on the web Codex — included with your plan, and it does everything
Pro Yes, in developer mode Reading only — it can look at your campaigns and stats, not change them Codex for the full set; the web connector for read-only questions
Business Yes — an owner turns on developer mode for the workspace Yes, with a confirmation step before each write action Either. The workspace connector for everyone, Codex for yourself
Enterprise / Edu Yes — an admin grants developer access, then chooses who may use the connector Yes, with a confirmation step before each write action Either

OpenAI documents the developer-mode switch at Settings → Security and login → Developer mode and lists it for Pro, Plus, Business, Enterprise and Education accounts on the web — but the full connector support that lets an agent change anything is the part reserved for the workspace plans, and write actions always ask you to confirm before they run. Details are in OpenAI's own developer mode guide. If your screen does not match what you read here, trust your screen — OpenAI moves this quickly, and Codex works regardless.

How do I set it up in Codex?

Codex is OpenAI's coding agent, included with every paid ChatGPT plan from Plus upward, and it speaks MCP natively. Two lines and you are connected.

  1. Create a WarmySender key. In WarmySender, go to Settings → API Keys and create one (it starts with ws_). Grant everything for full control, or only the read permissions if you want your agent to look but not touch. The key is shown once — copy it somewhere safe.
  2. Register WarmySender with Codex. In your terminal: 
    export WARMYSENDER_API_KEY=ws_YOUR_API_KEY_HERE
codex mcp add warmysender --url https://warmysender.com/mcp --bearer-token-env-var WARMYSENDER_API_KEY
    Add the first line to your shell profile so it survives a restart.
  3. Or skip the key entirely. Run codex mcp add warmysender --url https://warmysender.com/mcp and then codex mcp login warmysender, and sign in with your normal WarmySender account when the approval screen opens.
  4. Ask for something. "List my campaigns and tell me which ones sent nothing yesterday." If that answers, you are live.

Codex keeps its servers in a config file at ~/.codex/config.toml, so you can also add WarmySender by editing that file directly instead of running the command — the format and the codex mcp commands are documented in OpenAI's Codex MCP guide. In the Codex IDE extension, open its settings, choose MCP servers, add WarmySender the same way, then restart the extension so it picks the server up.

How do I set it up in a ChatGPT Business, Enterprise or Edu workspace?

  1. An owner or admin turns on developer mode for the workspace (Settings → Security and login), which is what makes custom connectors available at all.
  2. Add a custom connector pointing at https://warmysender.com/mcp. Choose the sign-in route if it is offered — then each person approves with their own WarmySender account and only ever sees their own workspace. Otherwise supply a ws_ key.
  3. On Enterprise and Edu, the admin then chooses who may use the connector, and can restrict what it is allowed to do before publishing it.
  4. Each member enables it in their own settings and starts a new chat. Write actions ask you to confirm before they run — read the summary, then approve.

If you are the person doing this for a team, the read-only route is a good first week: grant only read permissions on the WarmySender key, let everyone ask questions about campaigns and inbox health, and widen it once people are comfortable.

What about the ChatGPT desktop app?

The desktop app has its own MCP servers screen in Settings: add a server, give it a name, choose the streamable HTTP type, paste https://warmysender.com/mcp, save, then restart the app. Servers that use a sign-in show an Authenticate button — click it and approve with your WarmySender account. Because OpenAI is still rolling this out, what you see may differ from what you read anywhere; if the screen is not there yet, Codex is the reliable route today.

API key or sign-in — which should I use?

The two ways to authorise an agent, and when each is the better fit.
Sign in with your WarmySender account API key
Setup Point the agent at the address, approve on the screen that opens Create a key once, paste it into the agent
Best for Your own laptop, one person, no key to look after Terminals, servers, shared or headless setups where nobody can click an approval
Choosing permissions Full access to your own workspace You pick exactly which permissions the key carries, including read-only
Turning it off Revoke it in Settings → API Keys Delete the key in Settings → API Keys

Both give your agent exactly the same access to your own workspace, and both stay inside the same safe sending limits. Neither ever exposes your mailbox passwords, your social logins or your billing details.

Frequently asked questions

Can ChatGPT Plus add a custom MCP connector?

Not in a way that will run your campaigns. Developer mode — the switch that makes custom connectors possible — is listed for Plus on the web, but the full connector support that lets an agent create or launch anything is reserved for ChatGPT Business, Enterprise and Edu, and a personal Pro account is limited to reading. The good news is that your Plus plan already includes Codex, which connects to WarmySender in two lines and can do everything: build, launch, pause, resume, enroll, verify and report.

Why is a personal ChatGPT account read-only for custom connectors?

That is OpenAI's choice, not ours. Custom connectors are configured through developer mode, which OpenAI positions for people who understand what they are wiring up, and the full write-capable version is currently offered to workspace plans where an admin can review and approve what a connector may do. WarmySender works the same either way — the limit is on the ChatGPT side, so the answer is to use Codex or a workspace plan rather than anything you change in WarmySender.

How do I connect Codex to WarmySender?

Create a key in WarmySender under Settings → API Keys, then run two lines in your terminal: export the key as a variable, and run codex mcp add warmysender --url https://warmysender.com/mcp --bearer-token-env-var WARMYSENDER_API_KEY. Prefer not to handle a key? Add the server without the key flag and run codex mcp login warmysender to sign in with your WarmySender account instead. In the Codex IDE extension, add the same server under its MCP servers setting and restart the extension.

I connected it but no WarmySender tools appear. What now?

Three things cause almost every case. First, your ChatGPT plan may not allow write-capable custom connectors — use Codex instead. Second, in a Business or Enterprise workspace the connector has to be added by an owner and then switched on by you individually, so check your own settings after the owner adds it. Third, the key you used may carry read-only permissions, in which case reading works and everything else stays hidden; create a key with the permissions you need in Settings → API Keys. After any change, restart the app or start a fresh chat so it reloads the tool list.

Can ChatGPT send my cold emails?

No agent sends anything, and that is deliberate. Your agent builds the campaign, enrolls the people and starts it — and starting it simply hands it to WarmySender's scheduler, which paces every send inside safe daily caps and a gradual ramp. Ask it to blast a thousand people today and WarmySender still sends them out gently over time. Your agent also cannot raise a limit, and it cannot connect or disconnect a mailbox or a social account — that always stays with you in the app.

Does this cost extra?

No. Connecting an AI agent is included with any active WarmySender plan — there is no separate add-on to buy. You will need a paid ChatGPT plan for Codex, but that is between you and OpenAI.

Still stuck? Email [email protected] — we respond within 24 hours on business days.