MCP server

Seegnals runs a remote Model Context Protocol server. An assistant that speaks MCP over Streamable HTTP can read your campaigns, companies, replies, offers and mailboxes, add prospects, enrol them in a campaign, exclude addresses and manage event subscriptions, with the same rules and the same permissions as the REST API.

Every tool is a thin proxy over one REST endpoint on the same origin, called with the token you connected with. The server can do exactly what the REST API allows and nothing more: same validation, same workspace isolation, same rate limit.

Connecting

Claude Code

claude mcp add --transport http seegnals https://app.seegnals.com/api/mcp \
  --header "Authorization: Bearer sgn_your_token_here"

/mcp inside Claude Code then lists seegnals as connected with its tools.

Cursor

.cursor/mcp.json in the project, or ~/.cursor/mcp.json globally:

{
  "mcpServers": {
    "seegnals": {
      "url": "https://app.seegnals.com/api/mcp",
      "headers": { "Authorization": "Bearer sgn_your_token_here" }
    }
  }
}

Claude Desktop

The custom connector dialog in Claude Desktop expects OAuth. This server uses a bearer token, so bridge it with mcp-remote in claude_desktop_config.json:

{
  "mcpServers": {
    "seegnals": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://app.seegnals.com/api/mcp",
        "--header", "Authorization: Bearer sgn_your_token_here"
      ]
    }
  }
}

Any other client

{
  "mcpServers": {
    "seegnals": {
      "type": "http",
      "url": "https://app.seegnals.com/api/mcp",
      "headers": { "Authorization": "Bearer sgn_your_token_here" }
    }
  }
}

Raw protocol, for debugging. The Accept header is required by the transport:

curl -s https://app.seegnals.com/api/mcp \
  -H "Authorization: Bearer sgn_your_token_here" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

const res = await fetch("https://app.seegnals.com/api/mcp", {
  method: "GET",
  headers: {
    "Authorization": `Bearer ${process.env.SEEGNALS_TOKEN}`,
    "Content-Type": "application/json",
    "Accept": "application/json, text/event-stream"
  },
  body: JSON.stringify({
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/list"
  })
});
const json = await res.json().catch(() => null);
console.log(res.status, json);

import os, requests

res = requests.get(
    "https://app.seegnals.com/api/mcp",
    headers={"Authorization": f"Bearer {os.environ['SEEGNALS_TOKEN']}", "Accept": "application/json, text/event-stream"}, json={
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/list"
},
    timeout=20,
)
print(res.status_code, res.json() if res.content else None)

Because the server is stateless, initialize is not required before other calls and no Mcp-Session-Id is ever issued. Each request stands on its own.

Tools

Tools return the REST API’s JSON as text in content[0].text and, for objects, as structuredContent. Arguments are validated before any HTTP call is made.

Read tools are marked read-only in their annotations, so a client that asks before running write tools can run these without asking. Every tool mirrors the matching page in the REST reference; the parameters, filters and pagination are identical. The server never exposes more than the REST API does; when an endpoint ships, its tool is added the same day and listed in the changelog.

Errors

Tool failures are always tool results, never JSON-RPC protocol errors, so the model can read the message and correct its input.

Limits

• Rate limit. The REST budget is 60 requests a minute per token. One tool call spends two of them (the MCP request and the proxied REST call), so plan on about 30 tool calls a minute per token. tools/list costs one.
• Bodies. Prospects up to 64 KiB, suppressions and subscriptions up to 4 KiB, as in the REST API.
• Timeout. A proxied call is aborted after 20 seconds and reported as a tool error.
• Stateless. No sessions, no resumability, no server-initiated notifications, no resources or prompts capabilities.

Security notes

The token in your client configuration is a full-access workspace token. Keep the config file out of version control, give the assistant its own token so you can revoke it independently, and remember that an assistant with this server can create prospects and subscriptions in your workspace on its own initiative. Review what it did in Settings → Integrations → History and in the prospects list, where API-created prospects carry source = API.