MCP server
Connect your AI to EaseWeb
EaseWeb speaks the open Model Context Protocol (MCP), so any AI assistant that supports it can connect: Claude, ChatGPT, Cursor, Gemini, GitHub Copilot, Codex and more. Once connected, the assistant works with your presentations itself: builds them from a plan, edits slides, links them with buttons, reads other people's presentations and leaves notes on them.
https://easeweb.host/mcp/In most apps this address is all you need, with no key. The app sends you to EaseWeb: sign in (Google, Apple or Yandex) and choose the access: read and write or read only.
Chat assistants
Claude (claude.ai and the app)
- Settings → Connectors → Add custom connector.
- Name
EaseWeb, URLhttps://easeweb.host/mcp/, then Connect and sign in to EaseWeb. - In a chat, turn EaseWeb on in the tools menu.
ChatGPT
- Turn on developer mode: Settings → Apps → Advanced settings → Developer mode. It is not available on every plan.
- Settings → Apps → Create: name
EaseWeb, URLhttps://easeweb.host/mcp/, authentication OAuth. - Sign in to EaseWeb when ChatGPT asks, then pick the app in a chat.
Code editors and the terminal
Cursor
In ~/.cursor/mcp.json (or the project's .cursor/mcp.json):
{ "mcpServers": { "easeweb": { "url": "https://easeweb.host/mcp/" } } }
VS Code (GitHub Copilot)
In .vscode/mcp.json, then Start above the server:
{ "servers": { "easeweb": { "type": "http", "url": "https://easeweb.host/mcp/" } } }
Gemini CLI
gemini mcp add --transport http easeweb https://easeweb.host/mcp/
Then in Gemini CLI: /mcp auth easeweb.
Claude Code
claude mcp add --transport http easeweb https://easeweb.host/mcp/
Then in Claude Code: /mcp and sign in in the browser.
Codex (OpenAI)
codex mcp add easeweb --url https://easeweb.host/mcp/ codex mcp login easeweb
Any other AI
- An MCP client with OAuth connects to the same address and does the rest: the server supports discovery, dynamic client registration and PKCE.
- An MCP client without OAuth: create a key on the AI agents page and send it as
Authorization: Bearer <key>. - No MCP (your own agent, a script, a GPT with Actions): the same operations are available over REST with the same key, see the OpenAPI description. To let an AI figure it out itself, give it https://easeweb.host/llms.txt.
- Just reading a presentation needs no connection at all:
https://easeweb.host/api/v1/presentations/<address>/readreturns it as text.
What to ask
- "Create a presentation from these notes: a title slide, the plan, a slide per topic, a summary."
- "Read the presentation https://easeweb.host/movie/<address>/ and leave notes where it is unclear or wrong."
- "Turn the cards of the plan slide into buttons that lead to the topic slides, and add a 'back to the plan' button."
- "Add a slide with an SQL query that can be run right in the presentation."
Access and safety
- The AI acts on your behalf and sees only what you can open: your presentations, open and published ones, and those you were given access to.
- With read-only access the AI does not even see the tools that change anything.
- Every change is kept in the history, marked with the AI that made it. Any version can be restored.
- Connected apps and keys are listed on the AI agents page. A disconnected app loses access at once.
- How we handle data: privacy policy.
Tools
| Tool | What it does | Access |
|---|---|---|
whoami |
Who am I | read |
get_format |
Block format reference | read |
list_presentations |
List presentations | read |
get_presentation |
Get a presentation for editing | read |
read_presentation |
Read any presentation | read |
create_presentation |
Create a presentation | changes |
update_presentation |
Change title, description or theme | changes |
delete_presentation |
Delete a presentation | deletes |
get_slide |
Get one slide | read |
add_slide |
Add a slide | changes |
update_slide |
Change a slide | changes |
edit_blocks |
Add, change or delete blocks | changes |
delete_slide |
Delete a slide | deletes |
reorder_slides |
Reorder slides | changes |
upload_image |
Upload a picture | changes |
list_comments |
Read the notes | read |
add_comment |
Leave a note | changes |
resolve_comment |
Resolve a note | changes |
delete_comment |
Delete a note | deletes |
request_access |
Ask for access | changes |
list_access_requests |
Access requests | read |
answer_access_request |
Answer an access request | changes |
get_access |
Access settings | read |
set_access |
Change access | changes |
list_history |
History of changes | read |
get_version |
Content of a version | read |
restore_version |
Go back to a version | changes |
For developers
- MCP over Streamable HTTP (stateless). Authorization: OAuth 2.1 (RFC 9728, RFC 8414, Client ID Metadata Documents or dynamic client registration RFC 7591, PKCE S256, refresh tokens) or an API key. Metadata: /.well-known/oauth-protected-resource/mcp, /.well-known/oauth-authorization-server, server card: /.well-known/mcp/server-card.json.
- REST: OpenAPI. Guide for AI agents: llms.txt.
Support
Questions and bugs: saneev.svyatoslav@gmail.com.