Summary
Want to manage your Astra agents just by chatting with AI? This guide walks you through how to connect your Wati workspace to Claude or ChatGPT using the Model Context Protocol (MCP). Once set up, you can build, monitor, and improve your Astra agents - all from a simple conversation. No dashboard hopping. No code. Setup takes about 5 minutes.
What you can do after setup
Once everything is connected, your AI assistant becomes your control center. You can:
Build Astra agents just by describing what you need (in your language)
Check your agent's performance without opening the dashboard
Update your agent's tone, instructions, or escalation rules in seconds
Pull leads, conversations, or analytics into your chat with Claude or ChatGPT
Connect your agent with other tools like Shopify, HubSpot, Slack, Notion and more
You can do all of this without leaving the Claude or ChatGPT window you already use every day.
If you’re used to switching between tools and tabs, this will feel like a big upgrade.
Before you start
Let’s quickly make sure you have everything ready:
1. A Wati workspace with Astra enabled
Already using Wati with Astra? You’re good to go.
New to Astra? You can create a new Astra workspace here.
2. A supported Claude or ChatGPT plan
Custom MCP connectors aren’t available on free plans yet.
Supported plans:
Claude: Pro, Max, Team, Enterprise
ChatGPT: Plus, Pro, Business, Enterprise
You only need 1 paid user to get started, even if your whole team will use the agents.
3. About 5 minutes
Setup: ~2–3 minutes
First interaction: the rest of the time
4. Choose your AI assistant
You can use either Claude or ChatGPT - both work well.
Go with Claude if you want the most reliable MCP experience
Stick with ChatGPT if your team already uses it daily
Some teams connect both and just use whichever is open.
Instructions
How to set up in Wati MCP in Claude
Follow these steps to connect Wati MCP to your Claude AI:
1. Open Claude
Use either:
Desktop app (Download Claude at
claude.ai/download)Web version at
claude.ai
The desktop app is slightly more convenient if you want a dedicated workspace. Here's a screenshot of Claude's desktop app for your reference:
2. Open Connectors
Go to Customize → Connectors in the sidebar
You’ll see a list of available integrations.
3. Add Wati as a connector
Go to + then click on Add custom connector
Enter:
Name: Wati or Astra
Server URL:
https://astra-mcp.wati.io/mcp
Click Add
What you’re doing here is telling Claude where to find Wati.
4. Connect and authenticate
Click Connect next to Wati
Sign in to your Astra account
Click Authorize
This is a secure OAuth flow - your password is never shared.
5. Verify it worked
Open Claude, start a new chat and type: List my Astra agents
Grant Claude permission to use Wati MCP if asked
✅ If you see your agents (or an empty list), you’re all set
❌ If not, head to the Troubleshooting section below in this guide
How to set up in Wati MCP in ChatGPT
Prefer ChatGPT? Follow these steps:
1. Open ChatGPT
Web:
chat.openai.comDesktop app
2. Open Settings
Click your profile icon (bottom-left)
Select Settings
3. Enable connectors
Go to Apps → Advanced settings
Turn on Developer Mode (if needed)
This unlocks custom MCP connections.
4. Add Wati connector
Click Create app
Enter:
Icon: Use a custom icon if you'd like (optional)
Name: Wati or Astra
Description: Custom connector to build, monitor, and improve Astra agents.
MCP Server URL:
https://astra-mcp.wati.io/mcp
Authentication: OAuth (auto-detected)
Click Create
5. Sign in to Wati
Complete the login and approval flow.
6. Verify it worked
Start a new chat and type: List my Astra agents
(make sure Wati is enabled for that chat)
How to troubleshoot Wati MCP
Here are the most common issues and quick fixes:
1. Connectors option missing
Your plan doesn’t support MCP
Upgrade to a supported plan
2. No popup when connecting
Enable browser pop-ups
Try again
3. Wati login error
Sign in to Wati in another tab
Retry
4. "Unknown tool" error
Type: Re-load the Wati tools and try again
5. 401 Unauthorized
Session expired
Reconnect from Settings → Connectors
6. 403 Forbidden
Permission issue
Ask an admin for access
7. Assistant not using Wati
Make sure the connector is enabled for the chat
8. Slow responses
First request may take longer
It speeds up after that
Security and data handling
Here’s what’s happening behind the scenes when you're using Wati MCP:
Category | Details |
What gets shared |
|
What stays in Wati |
|
AI provider policies |
|
How to disconnect Wati MCP
To revoke access:
Open Wati dashboard
Go to Settings → Connected Apps
Click Revoke
Access is removed immediately.
Admin controls
If you manage a team, you can:
Control which AI tools can connect
Set role-based permissions
View audit logs
Restrict access by environment
Set token expiry rules
Frequently Asked Questions (FAQs)
General Usage
1. Do I need both Claude and ChatGPT?
No. You only need to choose one of the tools.
2. Will this replace the dashboard?
No. This solution complements the existing dashboard and does not replace it.
3. Can non-technical users use this?
Yes. Non-technical users can use this as long as they are comfortable with basic chat interactions.
4. Can agencies manage this?
Yes. Agencies can manage this by using dedicated accounts with limited access.
5. Will other AI tools be supported?
Yes. Additional AI tool integrations are planned.
Pricing and Availability
6. Is Wati MCP free?
Yes. Wati MCP is free for paid Wati users, with no usage cap during the launch period.
Platform Compatibility
7. Does this work on mobile?
It depends on the specific app. Desktop and web platforms are more reliable.
Support and Help
8. How can I get help if I need support?
You can contact support using the help widget in your Astra dashboard or the help widget available in the Astra Help Center.
Additionally, you can also check the system status at status.wati.io, or attend office hours held on Thursday at 17:00 IST / 09:00 EST.
Getting Started
9. What does it mean to be set up to run Wati from a conversation?
Being set up means you can now operate Wati directly through a conversational interface.