MCP server
Give Claude, ChatGPT, Cursor or any MCP-capable agent access to your TradeSgnl data, the Market Analyst and your signal webhook. One URL, sign in, allow.
Overview
TradeSgnl runs an MCP server. Add it to your agent with one address and it can read your signals, deliveries, accounts and performance, run the Market Analyst on any symbol your broker offers, and, if you allow it, send trade signals to your accounts.
The address is https://portal.tradesgnl.com/api/mcp. There is no key to copy: the agent opens the portal, you sign in and choose what it may do.
Chart control stays in the browser. An agent connected this way cannot draw on your TradingView chart, add indicators or manage TradingView alerts. For that, use the TradingView AI extension.
Connect
Add the address to your agent.
https://portal.tradesgnl.com/api/mcpClaude (web, desktop, mobile): Settings, Connectors, Add custom connector. Paste the address and save.
ChatGPT: Settings, Apps, Advanced settings, turn on Developer mode, then Create and paste the address.
Claude Code:
claude mcp add --transport http tradesgnl https://portal.tradesgnl.com/api/mcpCursor, VS Code, Windsurf and other MCP clients: add a remote server of type HTTP (sometimes called Streamable HTTP) and paste the address.
Sign in. The agent opens a TradeSgnl tab. If you are not signed in, sign in as usual.
Choose what it may do, then Allow. The page names the app and lists what it will read. If the app asked to send trade signals, an off-by-default switch lets you allow that, and you can cap the lot size per signal and restrict the symbols it may trade. Those limits are enforced before anything reaches your accounts. Click Allow and the tab closes.
That is it. The agent is connected until you disconnect it.
Manage connected apps
In the portal go to Profile, API access. Every connected app is listed with when it was last used and whether it can send signals. From there you can:
- Edit an app's trade permission and limits. The change applies to its next call.
- Disconnect an app. It stops working immediately; connect it again from the agent if you change your mind.
Connecting the same app twice replaces the earlier connection, so reconnecting is also the way to change your mind on a permission.
What the agent can do
| Tool | What it returns |
|---|---|
get_signals, get_signal_deliveries | The Signals page: every signal received and whether each account executed it |
get_clients, get_analytics, get_subscription | Connected accounts, balances, your plan |
get_trade_details, get_performance_metrics, get_performance_breakdowns, get_backtests | Positions, performance and backtests (Detailed Analytics where your plan includes it) |
get_alert_history, get_alert_preferences | Portal alerts |
list_licenses, list_signal_presets | Your licence codes and saved presets |
search_docs | This documentation |
list_accounts, get_symbols, get_candles, get_quote | Live prices from the account you trade on, as the Market Analyst reads them |
analyze_symbol | A full Market Analyst report. This is the only tool that spends AI credits |
send_signal | Sends a trade signal to your accounts. Only when you allowed it on the consent page |
send_signal takes a structured order (symbol, action, size, stop loss, take profit) and turns it into the same signal syntax TradingView alerts use. How much your agent checks with you before sending is up to you and how you set it up: an agent you have told to trade on its own trades; one you want to supervise can use dry_run, which returns the exact syntax and the accounts that would receive it without sending anything. After a send, the agent checks execution with get_signal_deliveries. Every signal an agent sends is tagged "AI agent" on your Signals page.
Limits
60 tool calls per minute, and for send_signal 10 per minute and 200 per day. The MCP Server is included from the Advanced plan.
Keys for scripts and tools
Anything that is not an agent, such as a spreadsheet, a WordPress table or your own script, cannot go through a sign-in page. For those, create a key under Profile, API access, and send it as Authorization: Bearer tsk_YOUR_KEY. The same key works against the MCP address and the REST API. A key carries the same trade permission and limits as a connected app, and is shown once, so copy it when you create it.
For Claude Desktop, Claude Code and Cursor, prefer connecting by address as above. A key is for tools that cannot open a browser.