OpenClaw Stock Alerts for Telegram, WhatsApp, and Discord

OpenClaw can connect an AI agent to Telegram, WhatsApp, Discord, and other channels. Stocksbrew can supply the stock research behind those conversations through MCP. The useful architecture is simple: OpenClaw owns the channel and schedule, while Stocksbrew owns the research, Radar state, change cursor, and alert rules.

This separation keeps channel credentials out of Stocksbrew and keeps financial permissions narrow. The agent can read research and manage a watchlist without receiving a brokerage connection or any ability to trade.

How the integration works

Telegram / WhatsApp / Discord
              ↓
           OpenClaw
              ↓
        Stocksbrew MCP
              ↓
 Research, Radar, changes, alerts

Stocksbrew does not directly send Telegram, WhatsApp, or Discord messages. OpenClaw routes the result through the channels you configure. Current Stocksbrew alert rules deliver by email. Chat updates come from a scheduled OpenClaw job that polls get_radar_changes.

Connect Stocksbrew MCP to OpenClaw

Add the remote MCP server, complete OAuth, and probe the connection. Check the current OpenClaw MCP documentation if your installed CLI differs.

openclaw mcp add stocksbrew   --url https://www.stocksbrew.online/api/mcp   --auth oauth

openclaw mcp login stocksbrew
openclaw mcp doctor stocksbrew --probe

During approval, choose the smallest useful scopes. A read-only morning summary needs stock, market, and Radar reads. Adding stocks or managing alert rules requires the corresponding write scopes.

Connection verification prompt

Show my Stocksbrew Radar and explain anything that changed today. Include the response timestamp and keep the returned cursor for the next check.

Connect your chat channel

OpenClaw supports channel-specific configuration. Follow its current setup for Telegram, WhatsApp, or Discord, then send a test message through that channel before adding the scheduled stock job. This verifies the outbound route independently from the financial data connection.

A clean setup proves two paths separately:

1. OpenClaw can call the Stocksbrew tools and receive current data.
2. OpenClaw can deliver a plain test message to the selected channel.

Only after both work should you schedule a combined Radar check. That makes failures easier to diagnose.

Schedule a useful Radar update

Check my Stocksbrew Radar every weekday before the US market opens. Use the saved cursor. Message me only when earnings, a directive, a risk, or a tracked price level changed. Keep the message under 10 lines. Save the new cursor only after the message is delivered.

The cursor prevents a daily agent from reading the entire watchlist as if everything were new. The delivery rule prevents notification fatigue. The agent should stay quiet when no material attention items exist.

Use Radar cursors correctly

get_radar_changes returns attention items and a cursor. Treat the cursor as opaque even if it currently resembles a timestamp.

{
  "attention_items": [
    {
      "item_id": "directive:NVDA:2026-08-01T10:30:00Z",
      "ticker": "NVDA",
      "kind": "directive_change",
      "occurred_at": "2026-08-01T10:30:00Z",
      "summary": "Call changed from Hold to Accumulate"
    }
  ],
  "cursor": "2026-08-01T10:30:00Z"
}

1. Call the tool with the last successfully delivered cursor.
2. Filter or summarize the unseen items.
3. Send the channel message.
4. Persist the new cursor only after delivery succeeds.
5. Back off on rate limits and server errors.

If delivery fails after the tool call, retaining the old cursor makes the next run replay the items. Stable item_id values provide a second deduplication key.

Handle a ticker with no detailed brief

OpenClaw can still answer with current baseline data. The response says whether detailed research exists and provides a machine-readable next action.

User: Why is CCL moving?

Agent: Stocksbrew has current price, technical, and fundamental context for CCL. Its detailed brief is not ready. Want me to add CCL to Radar and queue research?

The agent asks because adding a stock changes the member's Radar. After confirmation, add_to_radar queues or refreshes research. A later scheduled check can revisit the ticker when the coverage state moves from queued or running to ready.

Stock alerts and chat updates are different

A Stocksbrew alert is a price rule such as “NVDA below $190.” Current alert delivery is email. Earnings, RSI, call, anomaly, and news updates are produced by polling Radar changes and routing a summary to a channel.

OpenClaw can still manage alert rules conversationally:

User: Create an alert when NVDA drops below $190.

Agent: I can create a below-$190 email alert. Should I create it?

After confirmation, the agent calls create_alert. It can later list active rules with get_my_alerts or deactivate one with delete_alert.

Recommended OpenClaw permissions

• Channel summary only: stocks:read, market:read, and radar:read.
• Research assistant: add radar:write so confirmed additions can queue research.
• Alert assistant: add alerts:read and alerts:write.

Stocksbrew mutations are reversible and scoped. The MCP descriptions mark write and destructive operations. The agent should still ask before every change. Revoke the OAuth connection from Profile if a device or OpenClaw installation is no longer trusted.

Operational safety checklist

• Keep Telegram, WhatsApp, and Discord credentials inside OpenClaw's supported secret storage.
• Use OAuth for interactive Stocksbrew access. Never paste personal tokens into a chat.
• Send summaries only to destinations you control.
• Do not describe a queued brief as completed research.
• Do not convert research output into an automatic trade.
• Log the last successful cursor, not every failed attempt.

REST fallback

If a particular OpenClaw release cannot complete remote MCP OAuth, a deterministic integration can use the REST endpoints with a personal token stored as a secret. The same Radar, change, and alert operations are available. MCP remains the easier conversational option because the client can discover the tools and their schemas.