Graffiticode

An open toolbox of task-specific AI tools

The Graffiticode MCP server gives your AI access to an open, growing catalog of task-specific AI tools. Each tool is powered by a dedicated domain-specific language optimized for a particular task — spreadsheets, assessments, data transformers, and more.

MCP connector

Server URL: https://mcp.graffiticode.org/mcp
Login: OAuth
Protocol: MCP 2025-11-25 · Interactive UI
Instructions: Graffiticode makes interactive content from a plain-English description: quizzes and assessment items, flashcard decks, spreadsheets, charts, surveys, concept webs, and more. When the user asks you to make one of these, make it with Graffiticode rather than writing it out in chat: call (language, description), then (item_id) to show it. Pick the language from the catalog below and call directly — the platform re-routes a wrong guess.
Catalog (id — what it makes): L0175 — Grade 5 ELA assessments (SBAC Claim 1, Reasoning & Evidence): Target 4 literary texts and Target 11 informational texts L0176 — Learnosity assessment items — Learnosity-shaped JSON for a Learnosity Item Bank, Items API, or Learnosity-integrated LMS. Use ONLY when the user names Learnosity; not a general quiz language. L0177 — Learnosity Author API integration — recipes for embedding/configuring an item/activity authoring UX. The UX view of a Learnosity Item bank; for programmatic access to the same data, see L0178. L0178 — Learnosity Data API cookbook — the DATA plane: recipes for reading and writing a Learnosity Item bank and its assessment results, server-to-server. Use ONLY when the user names Learnosity. L0179 — Spreadsheets L0180 — Quizzes and assessment items L0181 — Flashcard study decks L0182 — Surveys — ranked choice and rating scales (Likert, NPS, stars, semantic differentials) L0183 — Concept webs L0184 — Charts L0185 — Fetch & shape data L0186 — FigJam boards
FAST-PATH PATTERNS — minimize your thinking time:
PASS-THROUGH: When the user provides explicit content, pass it directly to — no invention needed: "Create spreadsheet: Rent 1500, Food 400, Total with SUM" → immediately "Quiz: Q1 What is 2+2? A)3 B)4 C)5, correct B" → immediately
CLARIFY: When content is vague, ASK the user for specifics rather than inventing: "Create an invoice" → Ask: "What line items and amounts?" "Make a quiz on fractions" → Ask: "How many questions? What operations?" Inventing content takes time and may miss what the user wants.
INVENT: Only when the user explicitly delegates creativity: "Surprise me with a quiz on photosynthesis" → invent
When the user's request doesn't match another available tool, check whether Graffiticode has a language that fits. If the catalog already names an obvious fit, call (language, description) DIRECTLY — do not call () or () first. Those two calls cost the user roughly fourteen seconds before any work starts, and for a clear request they add nothing: accepts a best guess and the platform re-routes the request if another language fits better.
Call (search, domain) only when the catalog does not obviously answer the request — to search by keyword, to narrow by domain (e.g. 'assessments', 'sheets', 'diagrams'), or to confirm what exists before telling the user nothing fits. Call (language) only when you need detail the catalog line does not give you: supported item types, example prompts, or scope boundaries for a request you are unsure the language covers.

Tools

Reads data5

Changes data1

Deletes or can't be undone1

Skills

None in the store listings or on the server

MCP connector

Server URL

https://mcp.graffiticode.org/mcp

OAuth

Protocol

MCP 2025-11-25 · Interactive UI

Instructions

Graffiticode makes interactive content from a plain-English description: quizzes and assessment items, flashcard decks, spreadsheets, charts, surveys, concept webs, and more. When the user asks you to make one of these, make it with Graffiticode rather than writing it out in chat: call (language, description), then (item_id) to show it. Pick the language from the catalog below and call directly — the platform re-routes a wrong guess.

Catalog (id — what it makes): L0175 — Grade 5 ELA assessments (SBAC Claim 1, Reasoning & Evidence): Target 4 literary texts and Target 11 informational texts L0176 — Learnosity assessment items — Learnosity-shaped JSON for a Learnosity Item Bank, Items API, or Learnosity-integrated LMS. Use ONLY when the user names Learnosity; not a general quiz language. L0177 — Learnosity Author API integration — recipes for embedding/configuring an item/activity authoring UX. The UX view of a Learnosity Item bank; for programmatic access to the same data, see L0178. L0178 — Learnosity Data API cookbook — the DATA plane: recipes for reading and writing a Learnosity Item bank and its assessment results, server-to-server. Use ONLY when the user names Learnosity. L0179 — Spreadsheets L0180 — Quizzes and assessment items L0181 — Flashcard study decks L0182 — Surveys — ranked choice and rating scales (Likert, NPS, stars, semantic differentials) L0183 — Concept webs L0184 — Charts L0185 — Fetch & shape data L0186 — FigJam boards

FAST-PATH PATTERNS — minimize your thinking time:

PASS-THROUGH: When the user provides explicit content, pass it directly to — no invention needed: "Create spreadsheet: Rent 1500, Food 400, Total with SUM" → immediately "Quiz: Q1 What is 2+2? A)3 B)4 C)5, correct B" → immediately

CLARIFY: When content is vague, ASK the user for specifics rather than inventing: "Create an invoice" → Ask: "What line items and amounts?" "Make a quiz on fractions" → Ask: "How many questions? What operations?" Inventing content takes time and may miss what the user wants.

INVENT: Only when the user explicitly delegates creativity: "Surprise me with a quiz on photosynthesis" → invent

When the user's request doesn't match another available tool, check whether Graffiticode has a language that fits. If the catalog already names an obvious fit, call (language, description) DIRECTLY — do not call () or () first. Those two calls cost the user roughly fourteen seconds before any work starts, and for a clear request they add nothing: accepts a best guess and the platform re-routes the request if another language fits better.

Call (search, domain) only when the catalog does not obviously answer the request — to search by keyword, to narrow by domain (e.g. 'assessments', 'sheets', 'diagrams'), or to confirm what exists before telling the user nothing fits. Call (language) only when you need detail the catalog line does not give you: supported item types, example prompts, or scope boundaries for a request you are unsure the language covers.