Prompts.chat MCP with OpenCode.
Hello developer 👋! In this post we try the MCP of Prompts.chat, an open library of prompts for AI. The advantage is that we can get prompts that are optimized and structured by the community without leaving our editor. In this way we automate the writing of prompts and our tasks follow the same quality standard.

What is Prompts.chat
We are looking at the GitHub repository, which has around 176 thousand stars. In the About section it says:
f.k.a (Formerly Known As) Awesome ChatGPT Prompts. Share, discover and collect prompts from the community. Free and open source. You can also host it for your organization with complete privacy.
If we scroll down the documentation, the integrations section says that it can be used as an MCP. That means our AI agent can search the library of prompts directly from the editor.
Adding the MCP to OpenCode
We use the OpenCode AI agent and we want the Prompts.chat MCP available in every project, so we configure it globally. To do it we edit opencode.json and add the prompts-chat MCP in the mcp section:
"mcp":
Using the MCP with a React project
Now we open a personal project made with React. I open it in the Zed editor, but you can use any code editor. The idea is to ask for a good prompt from Prompts.chat to review the code of the project for best practices and see what can be improved.
I open a terminal in the center of Zed to start OpenCode. You can use any terminal where OpenCode works well; in some videos I also use Alacritty, one of the terminals that OpenCode recommends. We use the free model from the OpenCode cloud.
First we check that OpenCode has loaded the prompts.chat MCP. Then we put OpenCode in Build mode and ask:
Give me a prompt using the prompts-chat MCP to review the code of this React project and improve it in best practices. Save the prompt in the prompt.txt file, in the root of this project.
OpenCode searches for the right prompt through the MCP and saves it in prompt.txt. We can read the file to see what it did. From there we can edit the prompt, or ask OpenCode to run it directly and write the result of the code review in docs/code-review.md.
Conclusion
Prompts.chat gives us a large library of community prompts, and its MCP lets us use them without leaving the editor. With OpenCode we can configure it once, globally, and then ask the agent to find a prompt for a task such as reviewing a React project. We can save that prompt to a file, adjust it, or run it right away and write the result to a document. I leave it configured globally to use it whenever I need it in my projects.
Video
In the following video you can see the complete process (Spanish audio).