AI agents (MCP)
Connect Claude, ChatGPT, Cursor, Codex and other AI assistants to Webinaric over MCP so they can build webinars, publish them and read your leads.
On this page
Webinaric runs an MCP (Model Context Protocol) server. AI assistants that support remote MCP servers can connect to it and work in Webinaric as you. They can draft a webinar from a deck, write scripts, publish, and answer questions such as "Which campaign brought the hottest leads last month?". You don't need an API key. Every client signs in with your normal Webinaric login and asks for your approval.
Server URL
Open your avatar menu, choose Account, then Connected agents. Under Connect an agent, Server URL shows your server's address with a Copy button. It is:
https://webinaric.com/api/mcpThe server uses the Streamable HTTP transport and OAuth. Clients register themselves, so there is no client ID or secret to enter.
Connect your client
Add the server from your terminal. --scope user makes it available in every folder, not just the current one:
$ claude mcp add --transport http --scope user webinaric https://webinaric.com/api/mcpStart Claude Code and type /mcp. Pick webinaric (it shows "needs authentication") and choose Authenticate. Your browser opens Webinaric: sign in, check the request and click Allow. Back in the terminal, try "List my webinars in Webinaric."
What happens when you connect
- The agent registers itself. It tells Webinaric its name and where to send you back.
- You sign in. Webinaric opens in your browser. Sign in with an email code or Google, as you normally do.
- You approve. Webinaric shows the agent's name and what it will be able to do. Click Allow or Deny.
- The agent works as you. It can work in every workspace you're a member of, with the same access you have in the dashboard.
The agent stays signed in. If it goes unused for 30 days, you'll be asked to sign in again.
What agents can do
A connected agent can:
- See your organizations and workspaces.
- Read and edit webinars, slides, scripts, personas and knowledge bases. That includes poll, offer and Q&A slides, moderator settings, access codes and scheduled times.
- Import decks, generate slide contents and scripts, render audio and publish webinars. It can also list versions, republish an earlier one and restore a draft.
- Read analytics, the funnel, sessions and leads, and the questions attendees asked.
- Work the Inbox: suggest answers, answer questions by email, dismiss or reopen items, and add answers to a knowledge base.
- Manage workspaces and read the Agency reports, with the same plan and role rules as the dashboard:
| Tool | What it does | Who |
|---|---|---|
list_workspaces | Your organizations and workspaces, archived ones marked, and whether each organization has the Agency plan. | Everyone |
create_workspace | Creates an empty workspace. Needs the Agency plan. | Owners and admins |
rename_workspace | Renames a workspace. | Owners and admins |
archive_workspace | Archives a workspace: it is hidden and its public webinar links stop working. Sessions in progress finish, and nothing is deleted. The default workspace can't be archived. | Owners and admins |
unarchive_workspace | Brings an archived workspace back with everything in it. | Owners and admins |
usage_report | The usage report for a month: credits and USD per workspace. Needs the Agency plan. | Owners, admins and Member + Billing |
workspaces_overview | All workspaces analytics: totals and one row per workspace. Needs the Agency plan. | Everyone |
To import a deck or add a file to a knowledge base, the agent asks Webinaric for an upload link and uploads the file directly. Agents that can run commands, such as Claude Code, Cursor and Codex, can upload a PowerPoint, PDF, Word or Excel file from your computer. Files added to a knowledge base are converted to Markdown documents, the same as uploads in the dashboard.
Writing scripts, rendering audio and publishing use your plan's minutes, the same as in the dashboard. See Credits and billing.
What agents can't do
These stay in the dashboard:
- Generating intro avatar videos
- Importing from Google Slides
- Billing, members and organization settings (agents can read the usage report, but not buy credits or change the plan)
- Webhooks, tracking settings and email sender identities
- Voice previews
- Joining webinar sessions
Safety and limits
- Confirmation. Webinaric marks which tools only read and which publish, delete or email an attendee. Most clients ask you before the agent runs one of the latter.
- Attendee text. Questions and chat messages from attendees are labelled as untrusted for the agent, so it treats them as data, not instructions.
- Rate limits. Each user can make 120 calls a minute in total, and 30 a minute to tools that use AI or rendering. Over the limit, the agent is told to wait a minute and retry.
- Paged results. Long lists such as leads, sessions, questions and the Inbox come back 20 items at a time by default, up to 100, so they fit in the agent's context.
Disconnect an agent
Account → Connected agents lists every agent connected to your account, with when it connected. Agents on your computer show "Runs on your computer". Disconnect ends the agent's access on its next request. To use it again, connect and approve it again.