Connect an AI assistant (MCP)

eWebinar has an MCP (Model Context Protocol) server, so an AI assistant such as Claude, ChatGPT, Cursor or Claude Code can build and edit your eWebinars for you: settings, scheduling, the registration and thank-you pages, notifications, interactions, chat staffing, tracking scripts, integrations and series — plus the analytics and validation it needs to tell you why something isn't working.

Here's Claude building an eWebinar from a recording, start to finish:

It is a thin layer over the REST API v2. Every tool maps to a documented route and needs the same permission, so anything the assistant can do, you can do with curl. There is nothing to install and no local process to run.

Connect from Claude

From Home, choose New eWebinar (or the first-run create panel) and pick Create with Claude for a guided setup: copy your team's MCP connector URL, sign in and choose permissions in Claude, then open Claude with a starter prompt. You can also connect Claude yourself:

In Claude (web, desktop or mobile), open Customize → Connectors, choose Add custom connector, name it eWebinar and paste the server URL below. Claude sends you to eWebinar to sign in, you choose what to allow on the consent screen, and you are back in the chat. Ask it to "create an eWebinar from my recorded video" and let it lead.

On a Team or Enterprise plan, an owner adds the connector under Organization settings → Connectors first, and you then connect it from Customize → Connectors. Nothing else is needed — the server tells Claude how to sign you in.

https://api.ewebinar.com/mcp

Connect from ChatGPT

Building eWebinars from ChatGPT needs a ChatGPT Business, Enterprise or Edu workspace, on ChatGPT on the web. A Pro account can connect eWebinar read-only: enough to report on an eWebinar, but not to create or edit one. Plus and Free accounts can't add custom MCP servers.

On a Business workspace, only admins and owners can add a custom MCP server, so ask one to add eWebinar and publish it for the workspace first. On Enterprise or Edu, admins can also give selected members permission to add it. Once it's published, everyone in the workspace finds eWebinar under Settings → Plugins. To add it:

  1. In ChatGPT on the web, open Plugins, select +, then Add custom MCP server.
  2. Name it eWebinar, paste the server URL above under Connection, and choose OAuth for authentication.
  3. Review the warning, select I understand and want to continue, then Create as a plugin.
  4. Install the plugin, sign in to eWebinar when ChatGPT asks, and choose what to allow on the consent screen.
  5. In a new chat, type @, pick eWebinar and ask it to create an eWebinar from your recorded video.

Connect from other assistants

The server speaks Streamable HTTP at the same URL and advertises its OAuth configuration, so any MCP client that supports OAuth connects with the URL alone.

  • Claude Code — in a terminal:claude mcp add --transport http ewebinar https://api.ewebinar.com/mcp
  • Cursor — Settings → MCP → New MCP server, or in .cursor/mcp.json:
    {"mcpServers":
       {"ewebinar": {"type":"http","url":"https://api.ewebinar.com/mcp"}}
     }
  • Anything else — point it at the URL; your client performs the sign-in.

Account → Developers in eWebinar (and the API access integration under Team Settings → Integrations) generates these snippets for you, already filled in for your account.

What you allow

Signing in is the recommended way to connect. The assistant then acts as you: your name is on every change it makes, and you can withhold individual permissions on the consent screen. Which ones you see depends on your own role — you cannot grant an assistant something you cannot do yourself.

  • Read — see eWebinars, registrants, conversations and analytics. Always granted.
  • Edit drafts — change draft eWebinars, pages, notifications and interactions.
  • Publish — make a draft live, and change settings that apply immediately.
  • Integrations — connect and manage the team's integrations.
  • Registrants — read registrants and register attendees.
  • Delete registrants — delete a registrant.

What to expect

  • Drafts are safe. Edits stay on the draft until you publish, and every change is a revision the assistant can show you or revert. A good assistant summarizes what changed and asks before publishing.
  • Your recording goes in through eWebinar. Claude and ChatGPT can't upload a file from your computer, so the assistant sets up the draft, then asks you to upload your recording in the editor or share a direct download link. YouTube and Vimeo page links don't work. Claude Code and Cursor can usually upload a file from your computer themselves.
  • Some things are live immediately. Moderators and chat staffing, tracking scripts, integrations and team settings are not part of the draft, so changes to those take effect at once. The assistant is told to confirm first.
  • It can tell you why something isn't working. Ask why an eWebinar isn't getting registrations, or why an email didn't send, and it has the tools and the guidance to work through it.
  • Read-only tools run without asking; anything that publishes, overwrites or deletes asks for confirmation.

Starter prompts

Claude and other capable clients offer these from a menu before you have typed anything. They start a conversation; they are not tools.

  • create-ewebinar — Create an eWebinar. Interviews you about your goal, then designs and builds the eWebinar rather than asking which settings you want. Optional: goal.
  • improve-ewebinar — Improve an eWebinar. Reviews an existing eWebinar against best practice and reports what is worth changing, changing nothing until you agree. Optional: webinar.
  • set-up-crm-sync — Set up CRM sync. Connects a CRM or email tool and routes registrant and attendance data into it, including the tag routing that decides which eWebinars a hook applies to. Optional: integration, webinar.

Connecting without a browser

For a build, a cron job or a server — anywhere nobody can answer a sign-in prompt — use a team API key instead. Create one under Account → Developers, tick the same permissions as above, and send it as a bearer token:

Authorization: Bearer ew_api_...

A key belongs to the team rather than to a person, so changes made with one appear in revision history under the key's name. Keys can be revoked individually, and the secret is shown once. A key can publish just as a signed-in assistant can, a first publish included, unless publishing would charge you (see Publishing is refused below). A key cannot call GraphQL — REST v2 and the MCP server are the supported surfaces for machines.

Full reference

The complete prompt and tool reference — every starter prompt and every tool with the permission it needs, its arguments and the REST route behind it — is in the developer portal at api.ewebinar.com/docs/mcp (sign in with your eWebinar account to read it).

Troubleshooting

  • The client says it can't authenticate. Check the URL is exactly https://api.ewebinar.com/mcp, with no trailing path.
  • A tool returns 403. The connection was not granted that permission. Disconnect and reconnect, ticking the box you left unchecked; for a key, create a replacement with the permission ticked — an existing key cannot be widened.
  • Publishing is refused. An assistant can't publish when publishing would charge you, by ending your free trial or moving you to a higher subscription level, whether it signed in or uses a key. Publish that one from the eWebinar dashboard, where you see the charge before you confirm.
  • Reading the whole registrant list returns 429. Ask the assistant to sync what changed since last time rather than the whole list.