Treatwell MCP Server
# Treatwell MCP server
Connect Treatwell Connect to ChatGPT, Claude and any AI agent through Cotlet.dev. Read bookings and fill rate, prepare offers for quiet hours and manage your salon diary in your own signed-in account. Write actions ask for approval by default.
## What you can ask
> Summarize new bookings and cancellations since yesterday.
> Find quiet hours in next week's calendar and draft an offer for my approval.
> Read this month's revenue and staff performance reports.
## Connect in 2 minutes
Get your API key at https://cotlet.dev/app and export it as `COTLET_API_KEY`.
### ChatGPT
Settings → Apps & Connectors → Create, with connector URL `https://cotlet.dev/mcp/v1/connect`, then sign in with Cotlet.dev. No API key needed.
### Claude
Settings → Connectors → Add custom connector, with URL `https://cotlet.dev/mcp/v1/connect`, then sign in with Cotlet.dev.
### Claude Code
```bash
claude mcp add --transport http cotlet https://cotlet.dev/mcp/v1 --header "Authorization: Bearer $COTLET_API_KEY"
```
### Cursor
Copy `.cursor/mcp.json` from this repo into your project:
```json
{
"mcpServers": {
"cotlet": {
"url": "https://cotlet.dev/mcp/v1",
"headers": {
"Authorization": "Bearer ${COTLET_API_KEY}"
}
}
}
}
```
## Run it locally (stdio)
This repo is a real MCP server. It speaks MCP over stdio, lists its Treatwell tools without any key, and forwards every tool call to the hosted Cotlet.dev server. The hosted URL above stays the recommended way. Use this when your client only starts local servers.
Run the included Python module with Python 3.10 or later:
```bash
python3 treatwell_agent.py serve
```
Or with Docker:
```bash
docker build -t treatwell-mcp .
docker run -i --rm -e COTLET_API_KEY treatwell-mcp
```
Tools: `treatwell_guide`, `treatwell_status`, `treatwell_open`, `treatwell_read`, `treatwell_click`, `treatwell_type`, `treatwell_close`. Without `COTLET_API_KEY` a tool call answers: "Set COTLET_API_KEY. Get a key at https://cotlet.dev/app".
## Command line
Python 3.10+, with no third-party runtime dependencies. Clone this repository and run:
```bash
export COTLET_API_KEY="your-key"
python3 treatwell_agent.py connect # ready config for ChatGPT, Claude, Claude Code and Cursor
python3 treatwell_agent.py status # is Treatwell ready in your Cotlet.dev account
python3 treatwell_agent.py guide # Treatwell screens and sign-in link
python3 treatwell_agent.py read # open Treatwell, read the page, close the browser
python3 treatwell_agent.py tools # list Cotlet.dev tools
python3 treatwell_agent.py serve # start the stdio MCP server (same as no argument)
```
## Questions from Treatwell users
**Which Treatwell Connect login should I use?**
Use the login whose work the agent should do. The agent sees what that login's permissions allow.
**Do marketplace bookings appear?**
Treatwell marketplace bookings and your own clients share the Treatwell Connect calendar.
**Is Cotlet affiliated with Treatwell?**
No. Cotlet.dev is independent and not affiliated with Treatwell.
## Learn more
- [Treatwell integration and hosted tools](https://cotlet.dev/integrations/treatwell-api-mcp-server)
---
More: https://cotlet.dev/integrations/treatwell-api-mcp-server?utm_source=github&utm_medium=readme&utm_campaign=treatwell
Request access: https://cotlet.dev/signup?app=treatwell&utm_source=github&utm_medium=readme&utm_campaign=treatwell
## Workflows on the hosted integration
The live integration page documents these 12 operations:
- `view_calendar`
- `list_booking_changes`
- `manage_appointments`
- `block_time`
- `set_no_show_policy`
- `manage_clients`
- `import_clients`
- `record_payment`
- `create_promotion`
- `reply_to_review`
- `update_marketplace_profile`
- `view_reports`
The local stdio adapter exposes seven browser helpers listed above. These helpers forward to Cotlet.dev; the workflow operations are documented on the hosted integration page.
TDQS
Scored across 7 tools
Each tool maps to a distinct role: guide is documentation, status is an account/connection health check, and open/read/click/type/close form an unambiguous browser-automation loop. The descriptions explicitly sequence them (guide first, read before click/type, close when done), so an agent can pick the right tool without guessing.
All seven tools share the treatwell_ prefix followed by a single, consistent action verb (guide, status, open, read, click, type, close). There are no mixed conventions, casing inconsistencies, or redundant verb styles.
Seven tools is a tight, well-scoped set: two onboarding/health tools plus five browser primitives that each earn their place. Nothing appears redundant or padded, and the surface is large enough to complete real work.
The open/read/click/type/close loop covers navigation, inspection, and form interaction for the Treatwell web UI, with open accepting a target URL. Minor gaps exist around keyboard actions (Enter/submit shortcuts) and native selects, but those are largely workable by clicking numbered elements or submit buttons.