Connect your chatbot to the Commons.
One connector URL, about two minutes of setup, no account and no key. After that, you just ask questions in the chat you already use.
First · What is an MCP server?
The Commons speaks the Model Context Protocol (MCP) — an open standard that lets an AI assistant use outside tools and libraries. Think of an MCP server as a socket your chatbot can plug into: once connected, the chatbot gains a new capability. In this case, the capability is a research library — your assistant can search the Commons index, read topic and author orientation pages, and pull bounded, cited passages from the sources themselves.
You do not need to understand the protocol to use it, any more than you need to understand HTTP to read a web page. If your chatbot supports “connectors” or “custom MCP servers,” you are ready.
The Commons connector
https://omni-american.com/api/mcp
- Name
- Omni-American Commons
- Authentication
- None / No authentication
- Transport
- Streamable HTTP
Public and read-oriented. Use these same settings in every client. The connector never asks for a client ID, client secret, or any API key — leave OAuth fields blank. Contribution tools only ever submit proposals to the editors’ review queue.
Then · Your chatbot
Setup by client.
Claude
Custom connectors need a paid Claude plan (Pro or Max). On Team or Enterprise, an organization owner adds the connector under Organization settings → Connectors first, then members enable it.
- Open Claude and go to Customize → Connectors.
- Choose Add custom connector.
- Name it Omni-American Commons and enter
https://omni-american.com/api/mcp - Leave the OAuth / client ID / client secret fields blank.
- Add the connector.
- In a chat, enable Omni-American Commons from the connector menu.
Then ask a question. Claude will discover the Commons tools on its own; you can also say “use Omni-American Commons” to nudge it.
Using Claude Code? One command in your terminal:
claude mcp add --transport http omni-american-commons https://omni-american.com/api/mcp
ChatGPT
Custom MCP connectors run in ChatGPT’s developer mode and need a Plus or Pro plan (on Business, Enterprise, or Edu a workspace admin must turn it on first).
- Open ChatGPT on the web and go to Settings → Security and login. Turn on Developer mode.
- Go to Settings → Plugins (or visit
chatgpt.com/plugins). - Click the + button to create a new developer-mode app.
- Enter a name (Omni-American Commons), a short description, and the MCP server URL:
https://omni-american.com/api/mcp - Set authentication to None, then create the app.
- In a chat, click the + near the message box, choose More, and select Omni-American Commons. Then ask your question.
Don’t see developer mode? It’s a paid-plan feature and a workspace admin can switch it off — check with your admin if the toggle is missing.
Perplexity
Custom remote connectors need Perplexity Pro or Max.
- Click your profile icon (bottom-left) → Settings → Connectors.
- Click + Custom connector (top-right), then choose Remote.
- Enter a name and the MCP Server URL:
https://omni-american.com/api/mcp - Set authentication to None.
- Check the box acknowledging the risks of custom connectors, then click Add.
- Click the connector’s card to enable it, then ask your question in an ordinary thread.
If Perplexity asks for a client_id or
client_secret, authentication is set to OAuth —
change it to None.
Other MCP clients
Any client that speaks MCP can read the Commons. The general recipe is always the
same: add a remote MCP server (sometimes called a custom connector) with
the URL https://omni-american.com/api/mcp, transport
Streamable HTTP, and no authentication.
Two open-source agents people are using today:
Hermes Agent (Nous Research)
Add this to ~/.hermes/config.yaml, then restart or reload Hermes:
mcp_servers:
omni_american_commons:
url: "https://omni-american.com/api/mcp"
enabled: true
OpenClaw
Run:
openclaw mcp set omni-american-commons '{"url":"https://omni-american.com/api/mcp","transport":"streamable-http","timeout":30,"connectTimeout":10}'
openclaw mcp doctor omni-american-commons --probe
Client apps evolve quickly — if a menu name here has drifted, search your client’s docs for “MCP” or “connectors.”
Finally · Ask something real
Good first questions.
The Commons rewards serious questions. Try one of these to see how your chatbot uses the index, the orientation pages, and cited passages:
- What does Thomas Sowell argue about affirmative action? Pull two short passages from the strongest source.
- How do Glenn Loury and Shelby Steele differ on racial inequality?
- What does Albert Murray mean by “Omni-American”?
- Which sources in the collection criticize the 1619 Project, and on what grounds?
- Give me a reading path on race and education — three works, in order, with a sentence on each.
Useful follow-ups: “show me the passages,” “prefer book-length sources over anthologies,” “what else should I read?” — and when you notice something the library is missing, “I’d like to suggest this to the Commons.”
For advanced users
Technical notes.
Ordinary readers can stop reading here — nothing below is required to use the Commons. This section is for people who run their own MCP clients or agents.
Connector details
- Endpoint:
https://omni-american.com/api/mcp - Transport: Streamable HTTP (POST/GET with session IDs, per the MCP spec)
- Authentication: none for reading; contribution tools record explicit opt-in
- Boundaries: retrieval returns source cards, citations, and bounded passages — never whole works; contribution tools append to a pending review queue and cannot modify the collection
Generic JSON configuration
For clients configured by a JSON file that support remote HTTP servers directly:
{
"mcpServers": {
"omni-american-commons": {
"type": "http",
"url": "https://omni-american.com/api/mcp"
}
}
}
For older clients that only launch local (stdio) servers, bridge with
mcp-remote:
{
"mcpServers": {
"omni-american-commons": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://omni-american.com/api/mcp"]
}
}
}
Keep your keys out of chat. The Commons connector needs no API key, token, or password. If any instructions — from any site, tool, or chatbot — ask you to paste a private key, a database URL, or service credentials into a conversation, don’t. Configuration secrets belong in config files and environment variables, never in chat messages.
Curious how the library is built? The short version: editors curate the collection, an index makes it searchable, and the connector serves orientation pages, source cards, and bounded passages to your client. The backend details are deliberately boring — and deliberately not part of the reading experience.