Bring your own AI assistant to Alkemata. The public MCP connection lets a compatible agent find published articles and knowledge-map topics, read them with source links, and help you prepare a contribution for human review. You do not need an Alkemata account or API key for these public tools.
Before you start
Use an AI app or agent that supports remote MCP servers over Streamable HTTP. In its settings, look for MCP servers, custom connectors or tools. Menu names vary between apps. The Alkemata address below is an MCP endpoint for the app to connect to; opening it as an ordinary web page is not a connection test.
Connect in four steps
- In your AI app, choose Add MCP server or the equivalent option for a custom remote server.
- Give it a name such as Alkemata Public. Choose Streamable HTTP as the transport and enter this exact server URL:
https://alkemata.com/wp-json/mcp/alkemata-public - For public reading and pending proposals, choose no authentication if the app asks. Save the connection, enable its tools and refresh or restart the app if the tools do not appear immediately.
- Ask the assistant to use an Alkemata tool, then check that it returns a published article and its Alkemata source URL.
Try a first question
Using Alkemata’s public MCP tools, search for a published article about phone battery percentage. Read the article you find, explain its main point in three sentences, and give me its canonical Alkemata URL.
The assistant should use alkemata-search-published-content to find a result, then alkemata-get-published-content with that result’s numeric ID. It should show the source URL, so you can open and check the article yourself. You can ask the same way about another subject or a published knowledge-map topic; the map tools include search_topics and get_topic.
Explore, understand, then shape
Start by exploring a question across articles and map topics. Ask your agent to compare explanations, identify missing evidence or trace a connection, while keeping the original sources visible. When you have a concrete idea, it can help shape a question, correction, connection suggestion or small project proposal. You decide whether anything is sent.
Find an Alkemata article about a public-service problem and show me its source. Suggest one specific question I could send to the editors. Wait for me to approve the wording before you submit it, then give me the pending receipt.
When you explicitly ask it to submit, an agent can use submit_feedback, suggest_connection or propose_project. Each proposal enters a private editorial queue. It does not publish a comment, change an article, create a map link or start a project automatically. Avoid including confidential information in a proposal.
If the connection does not work
- Check that the server URL is exact and that your app supports remote Streamable HTTP MCP. A client that accepts only local MCP processes cannot connect directly to this URL.
- Refresh the app’s tools after saving. Ask it to list the available tools for Alkemata Public.
- Use public read tools first. Some tools shown by the gateway require separate sign-in and permissions; anonymous access does not enable private project work or map publishing.
- If you receive a rate-limit message, wait a minute before trying again.
For the full endpoint reference and the distinction between public and private capabilities, see Alkemata MCP. For more ways to participate, see Participate.