MCP

Point any MCP client at https://mcp.zernio.com/mcp and sign in with your Zernio account; the assistant can then call every API endpoint, from creating and scheduling posts to answering DMs, running ads and reading analytics. The server is hosted, so there is nothing to install, and it implements the Model Context Protocol so it works with Claude, Claude Code, ChatGPT, Codex, Cursor, Windsurf, VS Code and any other MCP client.

First call

Add the server to your client and sign in. In Claude Code that is 2 commands:

claude mcp add --transport http zernio https://mcp.zernio.com/mcp
claude /mcp

Then ask the assistant to list your connected accounts. It calls accounts_list, which answers with the platform, username and id of each one:

Found 2 connected account(s):

- instagram: acmecorp (ID: 66b2e19d8c3f5a7e9d0b1c2d)
- linkedin: acme (ID: 66b2e19d8c3f5a7e9d0b1c2e)

Setup has the configuration for Cursor, Codex, ChatGPT, Claude, VS Code and every other client, from a one-line command to a config file.

Autonomous agents

An agent has nobody at a browser to sign in, so it sends an API key and calls the tools itself. The server speaks Streamable HTTP, so plain HTTP works. This publishes a post with the posts_create tool; account_id names the account when the platform has more than one.

curl https://mcp.zernio.com/mcp \
  -X POST \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "Authorization: Bearer $ZERNIO_API_KEY" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "posts_create",
      "arguments": {
        "content": "Hello world. This is my first post from an agent.",
        "platform": "linkedin",
        "account_id": "66b2e19d8c3f5a7e9d0b1c2d",
        "publish_now": true
      }
    },
    "id": 1
  }'

Response (200), a JSON-RPC result whose content[0].text is the tool's reply:

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [
      { "type": "text", "text": "✅ Published to linkedin (@acme)\nPost ID: 65f1c0a9e2b5af0012ab34cd" }
    ]
  }
}

Omit publish_now and the tool schedules the post schedule_minutes from now (60 by default); pass is_draft: true to save a draft instead. When the platform has several accounts and you omit account_id, the reply is an error that lists the candidate ids: call accounts_list, then retry with one of them.

Create a dedicated key per agent so you can revoke it on its own; Setup has the rest.

Tools

20 core tools cover the common flows. The rest are generated from the OpenAPI spec, one per API endpoint, named {category}_{operation} (ad_campaigns_list_ad_campaigns for listAdCampaigns).

How it behaves

The server keeps about 50 tools visible and the rest behind search_tools

Your client sees the core tools plus the everyday posting, queue, validation, account-health, analytics and engagement tools, so a client that caps its tool list still shows everything it needs. The other endpoints cost no context until the assistant looks for one: search_tools and call_tool load the long tail on demand.

OAuth or an API key

Clients that support OAuth need only the server URL, and you can revoke that client from your dashboard. Clients without OAuth, and autonomous agents, send an API key as Authorization: Bearer $ZERNIO_API_KEY, as the request above does. Setup has both flows.

Related

• Setup: per-client configuration, local server, and fixes when it fails.
• Tools: core tool parameters and the tools/list catalog.
• API keys
• Python SDK: the package the MCP server ships in.
• SDKs