Prompts.chat MCP
https://github.com/f/prompts.chat-mcp- Category
- AI Agents
- Rank
- No. 2021Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- f
- GitHub
- 34 stars
- Date
About
MCP server that exposes the prompts.chat curated awesome-chatgpt-prompts collection to AI agents.
What it does
It lets an AI assistant browse, search, and retrieve reusable prompts through MCP. Searches accept keywords and optional type, category, or tag filters. Retrieved templates can receive variable values before use.
Why it's ranked here
A focused, low-friction utility with remote and local setup options across several MCP clients. Its narrow scope keeps the interface understandable, but the local package remains dependent on the hosted prompts.chat service, and its documented save capability is not implemented in the supplied source.
What's good
Search filters cover prompt type, category, and tag, while retrieval supports templated inputs. Users can choose hosted HTTP access or a local Node process. Optional query configuration can restrict the available prompt pool by users, categories, and tags.
Tradeoffs
The local server is a proxy, not an offline prompt library, so it still needs the upstream service. Error handling returns useful messages but exposes little recovery guidance. The documentation advertises saving prompts with authentication, while the supplied implementation registers only search and retrieval tools.
How to use it well
Use it when an MCP-enabled coding assistant needs quick discovery and reuse of existing prompt templates. Apply filters to reduce noisy searches, then retrieve a chosen template with its variables. It does not replace prompt authoring, evaluation, version control, or an offline prompt store.
Technical notes+
src/index.ts implements a TypeScript stdio MCP server that forwards JSON-RPC requests to https://prompts.chat/api/mcp, accepts JSON or SSE responses, and registers search_prompts plus get_prompt; it also forwards prompts/list and prompts/get. README.md documents save_prompt, but that registration is absent from the supplied source. package.json publishes the compiled build directory as an npm executable and depends on the MCP SDK plus Zod. server.json declares npm packaging with stdio transport and optional API-key and query environment variables. tsconfig.json enables strict TypeScript targeting ES2022.
Observed
- License
- ISC
- Primary language
- TypeScript
- Package
- Published on npm as @fkadev/prompts.chat-mcp with an executable entry point
- Interfaces
- MCP over local stdio, with documented hosted HTTP access
- Runtime
- Node.js 18 or newer is required for local installation
- Client support
- Setup instructions cover Cursor, Windsurf, VS Code, Claude Code, and Zed
- Implementation scope
- The supplied source registers search and retrieval tools; the documented save tool is absent
Read from README.md, package.json, src/index.ts, server.json, tsconfig.json.
What it can do
Browse curated ChatGPT prompts collection
Search query or category filter → List of relevant prompt templates
Retrieve specific prompt template
Prompt ID or name → Complete prompt text and instructions
Search prompts by category
Category name (e.g., writing, coding, business) → Filtered list of prompts in that category
Access prompt metadata
Prompt identifier → Prompt title, description, tags, and usage instructions
Expose prompts to AI agents via MCP protocol
MCP client connection → Structured prompt data accessible to connected AI agents
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.