Prompts.chat
Prompts.chat is a community library for discovering, sharing, and collecting AI prompts. The public library includes text and image prompts as well as agent skills. The publisher describes the project as free and open source, with self-hosting for teams that want their own instance.
Finding reusable prompts
The site organizes community examples for tasks such as email writing, image generation, and coding. A prompt is a starting instruction for an AI tool; its presence in the library does not establish that a model will produce the result described in it.
The API documentation provides keyword search with category, tag, and content-type filters. It also supports retrieving a specific prompt. Templates can contain variables, with clients that support elicitation asking users to supply the values before returning the completed prompt.
Prompts.chat exposes public prompts as native MCP prompts, allowing compatible clients to offer them through prompt pickers or slash commands. This gives a team a way to retrieve a reusable instruction in the tool where it already works.
Private prompts and agent skills
Most API discovery features work without authentication. Saving prompts or accessing private ones requires an API key. The documentation says saved prompts are private by default unless the account's default has changed.
The API also supports agent skills composed of multiple files, including a required SKILL.md, reference documents, and scripts. Public skills can be retrieved without authentication; private skills require a key. Read community instructions and scripts before choosing to use them in your own workflow.
Running an instance
The self-hosting guide lists Node.js 24, PostgreSQL, and npm as prerequisites. An instance can have its own branding, theme, and authentication providers, including GitHub, Google, Azure, or credentials. Configuration options include private prompts, categories, tags, and change requests.
The project uses a CC0 license, according to its documentation. Optional semantic search and AI generation require a configured model API key, so running those features may involve separate provider costs even though the library itself is free. Hosting and database requirements also remain part of operating your own instance.