word-guesser-gpt-mcp
Provides an interactive Word Guesser game for ChatGPT, with six guessing attempts, Wordle-style letter feedback, on-screen/physical keyboard input, local progress persistence, and English/Brazilian Portuguese language support.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@word-guesser-gpt-mcplet's play the five-letter word game"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Word Guesser — ChatGPT word game
Word Guesser is an interactive five-letter guessing game for ChatGPT. Players get six attempts, Wordle-style letter feedback, a physical/on-screen keyboard, and local progress persistence. The app defaults to English and selects Brazilian Portuguese when ChatGPT supplies BR in the coarse user-location hint. Players can switch languages at any time.
This is an interactive-decoupled ChatGPT app:
server/contains a stateless TypeScript MCP server.web/contains a React widget bundled into one CSP-friendly HTML resource.shared/contains game rules used by the server, widget fallback, and tests.docs/APP_REFERENCE.mdis the durable implementation/product reference.
Run locally
Requires Node.js ^20.19.0 or >=22.12.0.
npm install
npm run build
npm startThe server exposes:
docs:
http://localhost:8787/privacy:
http://localhost:8787/privacyterms:
http://localhost:8787/termshealth:
http://localhost:8787/healthMCP:
http://localhost:8787/mcp
For standalone widget development, run:
npm run dev:webThen open http://localhost:5173. Standalone mode bundles the same English and Brazilian Portuguese word lists used by the MCP server; ChatGPT mode validates guesses through the MCP tool.
Useful checks:
npm run typecheck
npm test
npm run build:webRelated MCP server: openai-mcp
Connect to ChatGPT
Build and start the server locally.
Expose port
8787through a public HTTPS tunnel, for examplengrok http 8787.In ChatGPT, enable Developer mode under Settings → Security and login.
Open Settings → Plugins (or
chatgpt.com/plugins), create a developer-mode app, and pastehttps://YOUR-TUNNEL.example/mcp.Add the app to a new chat and ask: “Let’s play Word Guesser.”
Refresh the app in its ChatGPT settings after changing MCP tools or metadata.
No OpenAI API key is required: ChatGPT is the host and the app only exposes MCP tools plus a widget.
Azure Functions deployment
The app uses Azure Functions' self-hosted MCP custom-handler preview on Linux Flex Consumption,
with Node 22 and managed identity storage access. The three top-level app resources are the Function App,
Flex Consumption plan and storage account; Application Insights and paid log ingestion are not configured. Infrastructure is in
infra/main.bicep; the current target is rg-word-guesser-brazilsouth in Brazil South.
GitHub Actions deploys pushes to main (including merged pull requests) through
the deployment workflow. It runs tests,
builds the production package, publishes to the existing Function App, and checks
the live MCP endpoint. Azure OIDC trust is configured for this repository's main
branch, and the workflow includes its public Azure identifiers. No GitHub secrets
or variables are required. See GitHub deployment setup
for the configured identity and the automated setup script.
Build a deployment package and install its production dependencies:
npm ci
npm run build:azure
npm ci --omit=dev --ignore-scripts --prefix dist/azureAfter signing into Azure CLI, validate and provision infra/main.bicep at subscription scope
with infra/main.parameters.json, then publish:
./scripts/publish-azure.ps1
node scripts/check-mcp.mjs https://word-guesser.mcp.11feed.comThe publish script creates a Linux-compatible ZIP, preserves the configured WIDGET_DOMAIN,
and uses Flex OneDeploy with an Azure CLI bearer token. It performs no remote build.
Use -WidgetDomain https://YOUR-BOUND-DOMAIN to change the widget origin; the hostname must
already be bound to that Function App. Without an existing setting, its actual Azure hostname is used.
Update its subscription, resource group and app parameters when using another target.
The production custom domain is word-guesser.mcp.11feed.com. Its existing Azure DNS zone is
11feed.com in cloud-shell-storage-eastus: CNAME word-guesser.mcp points to
func-word-guesser-6e5da.azurewebsites.net, and TXT asuid.word-guesser.mcp holds the app's
custom-domain verification ID. The domain binding and free site-scoped managed certificate
are configured separately in Azure; infra/main.parameters.json preserves the matching widget origin.
For a fresh target, validate domain ownership, add the hostname, issue the certificate and bind SNI HTTPS
before using the custom origin. Keep unrelated zone records intact. The existing Azure hostname remains available.
Use https://word-guesser.mcp.11feed.com/mcp in ChatGPT Developer Mode,
with no authentication. Public submission still needs domain verification, review assets,
dictionary licensing/provenance, and testing inside ChatGPT. Deployment validation and results
are recorded in .azure/deployment-plan.md.
Current package baseline
Versions were resolved from npm on 2026-07-21 and pinned exactly for reproducibility:
Package | Version |
| 1.29.0 |
| 1.7.4 |
| 0.2.2 |
React / React DOM | 19.2.8 |
Vite | 8.1.5 |
TypeScript | 7.0.2 |
Zod | 4.4.3 |
Tailwind CSS | 4.3.3 |
Official references
Security note
npm audit --omit=dev currently reports advisories in the latest pinned MCP/UI dependency trees. The app does not use Hono's affected serve-static path or lodash's affected template/unset APIs, and there is no non-breaking upstream fix available for the Apps SDK UI lodash dependency at this baseline. Re-run the audit when upgrading dependencies and replace the pinned versions only after the build and MCP checks pass.
This server cannot be deployed
Maintenance
Related MCP Connectors
MCP server for OpenAI API (chat completions, image generation, embeddings) via AceDataCloud
MCP server for AI dialogue using various LLM models via AceDataCloud
MCP server for GLM chat completions using Zhipu AI models via AceDataCloud
Official MCP server for Agentwork — delegate tasks to AI agents with human-in-the-loop
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceMCP Server for the Perplexity API.53 PyPI66MIT
- AlicenseNot gradedqualityBmaintenanceMCP server that bridges ChatGPT Plus/Pro to Claude Code, enabling chat, deep research, and image generation via your own account.223 PyPI51MIT
- AlicenseAqualityCmaintenanceAn MCP server that helps solve Wordle by suggesting words based on feedback and providing starting words.3GPL 3.0
- FlicenseNot gradedqualityCmaintenanceMCP server for OpenAI ChatGPT API. Enables chat completions, vision analysis, embeddings, and model listing through natural language.-