Wire Tradeics MCP servers into the agents your team already uses. Hosts share Streamable HTTP + bearer auth — only the config shape changes. Prefer enabling Docs, Platform, and AI as separate MCP entries.
Base fields (every host)
| Field | Docs | Platform | AI |
|---|---|---|---|
| Transport | Streamable HTTP | Streamable HTTP | Streamable HTTP |
| URL | https://mcp.tradeics.example/docs |
https://mcp.tradeics.example/platform |
https://mcp.tradeics.example/ai |
| Header | Authorization: Bearer <TOKEN> |
same | same |
Cursor
{
"mcpServers": {
"tradeics-docs": {
"url": "https://mcp.tradeics.example/docs",
"headers": {
"Authorization": "Bearer ${TRADEICS_MCP_TOKEN}"
}
},
"tradeics-platform": {
"url": "https://mcp.tradeics.example/platform",
"headers": {
"Authorization": "Bearer ${TRADEICS_MCP_TOKEN}"
}
},
"tradeics-ai": {
"url": "https://mcp.tradeics.example/ai",
"headers": {
"Authorization": "Bearer ${TRADEICS_MCP_TOKEN}"
}
}
}
}
Put tokens in Cursor secrets or environment — never commit them. After reconnect, ask: “List Tradeics MCP tools grouped by server.”
Also install Agent Skills so the agent knows when to call which server.
Claude Desktop
Add remote Streamable HTTP entries in claude_desktop_config.json for each Tradeics server URL with the same Authorization header. Restart, then verify tools appear per server.
ChatGPT / custom OpenAI agents
- Create one MCP connector per Tradeics server.
- Attach bearer auth.
- Enable read tools first; gate write tools by policy.
Hosts without custom headers may use ?token= — rotate faster if you do.
Custom agents (SDK)
# Pseudocode — one session per MCP server
servers = {
"docs": "https://mcp.tradeics.example/docs",
"platform": "https://mcp.tradeics.example/platform",
"ai": "https://mcp.tradeics.example/ai",
}
headers = {"Authorization": f"Bearer {TRADEICS_MCP_TOKEN}"}
# for name, url in servers.items():
# session = await client.connect(url=url, headers=headers)
# tools[name] = await session.list_tools()
Safety checklist
- Start with Docs-only; add Platform reads; then writes.
- Keep tokens out of transcripts and public repos.
- Apply rate limits and errors policies.
- Prefer staging credentials when experimenting with writes.
Stuck connecting? Chat with support — include host name and which server fails.