MCP Server

One endpoint that lets an AI client work with every RichlyVa site you have access to — your CRM, task boards, invoices, short links, uptime monitors, HR records and more. It speaks the Model Context Protocol over Streamable HTTP (JSON-RPC 2.0).

Connect in one step

Most modern clients (Claude, Cursor, and anything else that supports remote MCP) only need the URL above. Paste it, and the client walks you through sign-in automatically — no tokens to copy, no config files to edit.

1. Add a custom connector / MCP server in your client and paste the URL.
2. Your browser opens a RichlyVa consent page. If you are not signed in yet, sign in first — you land back on the same page. Tick the sites and capabilities you want the client to use.
3. Done. The sites you ticked are connected for you automatically. If one says you have no account there yet, sign in to that site once and click Connect again — see Troubleshooting below.

Read-only capabilities are pre-ticked. Anything that writes is never ticked for you — you have to enable it deliberately. Sites you have not subscribed to never appear at all.

Tick only the sites this client actually needs. Some clients and models cap how many tools they accept (OpenAI models take at most 128 per request), and ten or so sites already come to a few hundred tools. If your client is one of those, use the discovery address below instead.

Client examples

Claude Desktop / Claude.ai

Settings → Connectors → Add custom connector → paste the URL. Sign in when the browser opens.

Claude Code

claude mcp add --transport http richlyva https://www.richlyva.com/api/mcp

Then type /mcp inside Claude Code, pick richlyva and choose Authenticate.

Cursor

Add this to ~/.cursor/mcp.json, then open Settings → MCP and click Connect next to richlyva:

{
  "mcpServers": {
    "richlyva": {
      "url": "https://www.richlyva.com/api/mcp"
    }
  }
}

VS Code (GitHub Copilot)

Add this to .vscode/mcp.json (or run MCP: Open User Configuration), then click Start above the entry and sign in when asked:

{
  "servers": {
    "richlyva": {
      "type": "http",
      "url": "https://www.richlyva.com/api/mcp"
    }
  }
}

ChatGPT

Needs a paid plan and works on the web only. Settings → Apps → Advanced settings → turn on Developer mode, then Create app. Paste the discovery address below (ChatGPT caps the number of tools), leave Authentication on OAuth, and click Create. On Business / Enterprise workspaces an admin has to allow developer mode first.

Codex CLI

codex mcp add richlyva --url https://www.richlyva.com/api/mcp/discover
codex mcp login richlyva

Clients that only support local (stdio) servers

Bridge with mcp-remote:

{
  "mcpServers": {
    "richlyva": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://www.richlyva.com/api/mcp"]
    }
  }
}

Clients with a tool limit (ChatGPT and other OpenAI-model clients, Codex, Cursor CLI…)

Use this address instead of the one above:

https://www.richlyva.com/api/mcp/discover

Instead of hundreds of tools at once, the model gets a few entry points: search_tools to find the tool it needs and call_tool to run it. Sign-in, the sites you tick and the confirmations for write actions are exactly the same.

Your own code

If you are scripting against the endpoint rather than using a client, create an access token under Account settings → AI access tokens and send it as a bearer token:

curl -X POST https://www.richlyva.com/api/mcp \
  -H "Authorization: Bearer rvpat_your_token_here" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

What you get

The tool list is built per user. You see tools for the sites you actually have, and only the capabilities you granted. The more sites you have, the more tools you get: around 300 with ten or so sites, and around 700 on the full matrix.

Tools are named site_action — for example hrm_listEmployees, tasks_createTask, invoice_listInvoices. Not sure where to start? Call richlyva_listMySites to see which sites you have. Where a site offers an overview tool, call it first to get oriented — the name is either _summary or _getSummary depending on the site (hrm_summary, invoice_getSummary).

What is never exposed

• Credentials. API keys, passwords, tokens and session state are stripped from every response, by field name, before anything leaves the server.
• Other people's data. Every tool recomputes who you are on the target site. Workspace and tenant boundaries are enforced per call, not inherited from a framework scope.
• Money and your account. No tool takes or sends a payment, touches your RichlyVa plans or billing, or grants credits, and none can change your account's email, role or permissions. (Prices of products in your own store or books are business data — with write access granted, those can be edited.)
• Deletion. Only a handful of tools delete anything, and every one asks for confirmation first: your own Aikeedo assistants and library items, a Form Builder folder (the forms inside are kept and become unfiled), an Eraseo history entry (a soft delete — the files stay), and a StackPosts draft that has not been sent yet. Nothing else is deleted. Bulk sending is not exposed either.

Troubleshooting

You can revoke any client at any time under Account settings → AI connections. Revocation takes effect on the next call.