How to Connect the WIN System API to n8n, OpenClaw & Agentic AI Services
The Bridge Between Desktop Intelligence and Automation
The WIN System is an observational tool — it watches, listens, and analyzes, but it does not take actions. That's by design. However, the intelligence it produces (transcripts, AI summaries, screenshots) is incredibly valuable as input to agentic AI systems that do take actions.
This is where the WIN System API comes in. When the desktop app is running, it exposes a local HTTP bridge on
localhost:5010 that any automation platform can call.
Available API Endpoints
| Method | Endpoint | Description |
|---|---|---|
| GET | /health |
Check if the bridge is running |
| POST | /win/start |
Start audio recording |
| POST | /win/stop |
Stop audio recording |
| POST | /win/ask |
Trigger AI analysis of transcript + screenshot |
| POST | /win/clear |
Clear the current transcript buffer |
| GET | /win/transcript |
Retrieve the current transcript text |
Connecting to n8n
n8n is an open-source workflow automation tool. Here's how to create a workflow that automatically captures a meeting transcript and sends it to Slack every 15 minutes.
Step 1: Trigger
Use an Interval node set to fire every 15 minutes (or use a Webhook node for on-demand triggers).
Step 2: HTTP Request — Get Transcript
Method: GET URL: http://localhost:5010/win/transcript Response Format: JSON
This returns the current transcript buffer as plain text.
Step 3: HTTP Request — Ask the AI
Method: POST
URL: http://localhost:5010/win/ask
Body: { "prompt": "Summarize the key points from this meeting in 5 bullet points." }
Response Format: JSON
This triggers a full AI analysis (transcript + screenshot) and returns the AI's response.
Step 4: Send to Slack / Email / CRM
Use n8n's Slack, Gmail, Salesforce, or HubSpot nodes to send the AI summary wherever you need it.
Connecting to OpenClaw Agents
OpenClaw agents operate inside browser sandboxes — they can navigate web pages, fill forms, and extract data. But they cannot hear audio or see your full desktop. The WIN System fills this gap.
Architecture Pattern
- WIN System captures the audio and screen context (the "ears and eyes").
- Your orchestration layer (n8n, a Python script, or a custom agent) calls
/win/transcriptto get the conversation. - OpenClaw agent receives the transcript as context and takes action in the browser — creating a CRM record, updating a project tracker, or sending a structured email.
Example: Auto-Create a CRM Lead After a Call
1. WIN System records the sales call 2. n8n polls /win/transcript when recording stops 3. n8n sends transcript to GPT / Groq: "Extract lead info as JSON" 4. OpenClaw agent opens Salesforce, fills in the Lead form with the extracted data 5. Lead is created without manual data entry
Connecting to Zapier
The WIN System's built-in Zapier webhook support lets you fire a webhook after every "Ask the AI" call. Configure the webhook URL in the Settings tab, and the AI response is automatically sent to your Zap. From there, Zapier can route it to 5,000+ apps.
Python / cURL Quick Start
For developers building custom integrations:
# Check health curl http://localhost:5010/health # Start recording curl -X POST http://localhost:5010/win/start # Get transcript curl http://localhost:5010/win/transcript # Ask the AI curl -X POST http://localhost:5010/win/ask # Stop recording curl -X POST http://localhost:5010/win/stop
Security Considerations
- The API only listens on
localhost— it is not accessible from other machines on the network. - No authentication is required because the API is local-only. If you need remote access, use an SSH tunnel.
- The WIN System itself is read-only — API calls can retrieve data and trigger analysis, but cannot modify files or take actions on your computer.
Conclusion
The WIN System is not a replacement for agentic AI — it's the sensory layer that feeds them. By combining WIN System's desktop-level awareness (audio, microphone, screenshots) with the action-taking capabilities of n8n, OpenClaw, or custom scripts, you build automation pipelines that actually understand what's happening in your real-world conversations.
See the full API reference
API Docs