eightytwenty

A free, read-only connector that gives your AI assistant a slide-design playbook for case competition decks, plus search over a curated library of slide layouts.

Each library entry is an abstract layout record (region boxes, colour and type roles, hierarchy, a short note on why the slide works) and a wireframe drawn from that record. There are no original slides, names or numbers in it. Your assistant uses a reference as inspiration and builds the slide with its own tools.

No account, no sign-in. It works with assistants that connect to remote MCP servers, such as claude.ai, ChatGPT, Claude Code and Cursor; web pages cannot call it directly.

Connect

Server URL (Streamable HTTP, no authentication):

https://mcp.eightytwenty.tech/mcp

claude.ai

Customize → Connectors → + Add → Add custom connector. Enter a name and the URL above, choose No sign in, and add it. In a chat, turn it on under + → Connectors, then pick its start prompt there or just ask for help with your deck.

Claude Code not yet tested

claude mcp add --transport http eightytwenty https://mcp.eightytwenty.tech/mcp

Then run /eightytwenty:start, or just ask for help with your deck.

Other MCP clients not yet tested

ChatGPT (developer mode), Cursor, VS Code, Gemini CLI and others: add a remote MCP server with the URL above and no authentication. For clients that use an mcpServers JSON file, such as Cursor:

{ "mcpServers": { "eightytwenty": { "url": "https://mcp.eightytwenty.tech/mcp" } } }

Tools

All five are read-only. There is no way to upload to the library or list all of it.

get_playbook(stage?): The methodology: workflow, narrative and impact slides, slide archetypes, anatomy and taglines, distinctive design, visual rules and design options, writing, a QA checklist, and how to audit the style of a deck you attach in chat. With no stage it returns the index and the workflow. Your assistant should call it before planning a deck.
find_references(purpose, slide_type?, industry?, data_type?, impact?, k?): Searches the library by what a slide must do, with optional filters; impact (none, build, spotlight, shade_ribbon, repeat) finds slides that reference decks repeat for effect, and few exist. Returns up to eight short summaries (k, default 8), at most one per deck: id, slide type, a score, purpose, why it works, the layout's component kinds and its impact device if any, plus the library version. When planning a whole deck, use it after the slide-by-slide storyline is confirmed.
find_deck_flows(purpose, industry?, length?): Searches the library's deck flows (how strong decks order their sections) by what a whole deck must achieve, optionally by industry and length (short, medium, long). Returns up to three summaries: deck id, arc, length, industry, section sequence, why the flow works, and a score. Use it after the slide-by-slide storyline is confirmed.
get_reference(id): One slide: its labels (purpose, why it works, archetype, insight device, impact), its component layout, its design-DNA record (the 40 largest regions), and its wireframe image where the client shows images. Given a deck id, it returns that deck's flow: its steps in order, each with a slide id, its impact device, and whether that slide is in the library.
find_images(kind, query, k?): Sources a logo, icon or photo for a slide being built: public-domain logo SVGs from Wikimedia Commons and Simple Icons, permissive icon SVGs from Iconify, and CC0, public-domain and CC BY photos from Openverse. Returns up to four results (k, default 3, or 4 for icons) with their licence and either the SVG text or the photo's download and page links, plus a credit line only when the licence requires one; photos come with thumbnails. The search terms are sent to those three services; see Privacy.

Try asking

“Use eightytwenty to plan the executive summary slide for my case on a regional grocery chain.”
“Find reference layouts for a slide that compares three market-entry options, and build one in my style.”
“Here is my deck. Audit its style with the eightytwenty playbook and fix slide 4.”

Limits

Each client IP address can make 20 tool calls a minute. Going over starts a cooldown of 1 minute that doubles for repeat bursts, up to an hour. Requests from claude.ai share one larger allowance, and searches across all users share a daily budget that resets at 00:00 UTC. Photo searches have their own daily budget; when it runs out, logos and icons still work.