TeamDesk MCP Server
MCP server for TeamDesk databases, enabling CRUD operations, search, and document generation via natural language.
README
TeamDesk MCP Server
MCP (Model Context Protocol) server for TeamDesk databases. Connects Claude (web, Desktop, and mobile), Claude Code, or any MCP-compatible client to the TeamDesk REST API v2.
Features
- 11 tools covering all TeamDesk CRUD operations + search + document generation
- Works with any TeamDesk database — just provide your token and database ID
- Two deployment modes:
- Local (
server.py) — runs on user's machine via stdio - Remote (
deploy/server_sse.py) — multi-user server with OAuth 2.0, Streamable HTTP, rate limiting, and caching
- Local (
- Correct URL encoding for accented table/column names (Portuguese, Spanish, etc.)
- Timeout handling with automatic retry (55s timeout, 2 retries with backoff)
- Filter injection prevention via input sanitization
- Stderr logging for request timing and debugging
Requirements
- Python 3.10+
- A TeamDesk REST API token (generate at: TeamDesk > Setup > Integration > REST API > Tokens)
- Your TeamDesk Database ID (visible in the TeamDesk URL:
teamdesk.net/secure/db/XXXXX)
Installation
Option 1: Remote Connector (Recommended)
No local installation required. Connect directly from claude.ai — works on web, Desktop, and mobile automatically.
When you add a custom connector on claude.ai, it syncs to Claude Desktop and mobile automatically. One setup covers all platforms.
- Ask your admin to create an access key (see Remote Deployment)
- Go to claude.ai > Settings > Integrations > Add Custom Connector
- Enter a name (e.g.,
TeamDesk) and the server URL:https://your-server.com/m/{your_key}/sse - Click Configure > approve access
- Done — works on web, Desktop, and mobile
Option 2: Local Installer (Windows)
- Download
install.bat,server.py, andrequirements.txtfrom this repo - Place them in the same folder
- Run
install.bat - Follow the prompts (it will ask for your token and database ID)
- Restart Claude Desktop
Option 3: Local Installer (Linux / Mac)
- Clone and run:
git clone https://github.com/Danielbluz/teamdesk-mcp-v2.git
cd teamdesk-mcp-v2
bash install.sh
- Follow the prompts
- Restart Claude Desktop
Option 4: Manual Setup
git clone https://github.com/Danielbluz/teamdesk-mcp-v2.git
cd teamdesk-mcp-v2
pip install -r requirements.txt
Then add to your client config (see next section).
Configuration
Claude Desktop (local mode)
Edit claude_desktop_config.json:
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - Mac:
~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"teamdesk": {
"command": "python",
"args": ["C:\\path\\to\\server.py"],
"env": {
"TEAMDESK_API_TOKEN": "your_token_here",
"TEAMDESK_DATABASE_ID": "your_database_id"
}
}
}
}
Claude Code
Add to .claude/.mcp.json in your project or ~/.claude/.mcp.json globally:
{
"mcpServers": {
"teamdesk": {
"command": "python",
"args": ["C:\\path\\to\\server.py"],
"env": {
"TEAMDESK_API_TOKEN": "your_token_here",
"TEAMDESK_DATABASE_ID": "your_database_id"
}
}
}
}
Alternative: .env file
Instead of setting env vars in the config, you can create a .env file next to server.py:
cp .env.example .env
# Edit .env with your credentials
Tools
| Tool | Description |
|---|---|
list_tables |
List all tables in the database |
describe_table |
Get table structure (columns, types, properties) |
get_records |
Query records with filters, sort, and pagination |
get_record |
Retrieve a single record by ID |
select_view |
Query records from a pre-configured View |
create_record |
Create a new record |
update_record |
Update an existing record by ID |
delete_record |
Delete a record by ID |
search_records |
Full-text search across text columns (auto-detected) |
upsert_records |
Insert or update records (match column must be Unique) |
gerar_documento |
Generate a DOCX from a TeamDesk Documents template |
Filter Examples
[Status] = 'Ativo'
[Amount] > 1000
Contains([Name], 'Solar')
[Date] >= ToDate('2026-01-01')
Sort Syntax
Column -- ascending (default)
Column//DESC -- descending
TeamDesk API Quirks
These are undocumented behaviors that this server handles correctly:
| Quirk | Detail |
|---|---|
| DELETE uses GET | {table}/delete.json?id={id} is a GET request |
| UPDATE uses POST | Not PUT. Body must include @row.id |
| Body must be array | All write operations require [{...}], not {...} |
| Upsert match must be Unique | Match column must be marked Unique in table setup, otherwise error 3106 |
Document endpoint has no .json |
Use /document?id=X, not /document.json?id=X |
Sort uses // |
Column//DESC, not -Column or Column DESC |
| Accented names are valid | Geração, Irradiação — must be URL-encoded, not stripped |
| No LIKE operator | Use Contains([field], 'value') (case-insensitive) |
Date literals use # |
[Date] >= #2026-01-01# or ToDate('2026-01-01') |
| Retrieve by ID | Use retrieve.json?id=X, not {id}.json |
Troubleshooting
MCP not appearing in Claude Desktop
- Make sure you restarted Claude Desktop completely (check system tray)
- Verify
pythonis in your PATH:python --version - Check the config JSON syntax (no trailing commas)
Slow responses
- TeamDesk servers can be slow during US business hours (15:00–18:00 UTC)
- The server retries automatically on timeout (up to 2 times)
- Check logs: Claude Desktop > Ctrl+Shift+I > Console > filter
[TeamDesk MCP]
Token issues
- Generate a new token at: TeamDesk > Setup > Integration > REST API > Tokens
- The token should be a 32-character hex string
- Both
TEAMDESK_API_TOKENandTEAMDESK_TOKEN(legacy) are accepted
Remote Deployment
The deploy/ folder contains server_sse.py — a multi-user remote MCP server designed for deployment on a VPS behind Nginx/TLS. It supports:
- OAuth 2.0 with auto-approve (PKCE S256) — required by claude.ai custom connectors
- Streamable HTTP transport (MCP SDK 1.26.0+) — the protocol claude.ai uses
- SSE transport — for Claude Desktop local config via URL
- Multi-user authentication via API keys stored in a TeamDesk table
- Rate limiting and response caching
- Per-user tokens — each user's API key maps to their own TeamDesk token
How it works
-
Admin creates a record in the
Acessotable (TeamDesk) with:Chave_MCP: unique key for the user (e.g.,john_company_2026)Token: the user's TeamDesk API tokenAtivo_MCP:Sim(to enable access)
-
Each user gets a personal endpoint:
https://your-server.com/m/{chave}/sse -
The server validates the key, retrieves the user's token, and proxies all MCP tool calls to the TeamDesk API using that token.
OAuth 2.0 endpoints
claude.ai requires these OAuth endpoints for custom connectors:
| Endpoint | Purpose |
|---|---|
/.well-known/oauth-protected-resource |
RFC 9728 — resource metadata |
/.well-known/oauth-authorization-server |
RFC 8414 — auth server metadata |
/register |
RFC 7591 — dynamic client registration |
/authorize |
Authorization endpoint (auto-approve) |
/token |
Token exchange (PKCE S256 verification) |
Technical notes
- claude.ai uses POST (Streamable HTTP), not GET (SSE) — the server handles both
- The server injects the
Accept: application/json, text/event-streamheader if missing (claude.ai sometimes omits it, causing 406 from MCP SDK) - Sessions are stateful (
stateless=False) — required for persistentMcp-Session-Id - Configure via environment variables (see
deploy/docker-compose.yml)
Docker deployment
cd deploy
cp .env.example .env
# Edit .env with your TEAMDESK_DATABASE_ID, TEAMDESK_MASTER_TOKEN, MCP_PUBLIC_URL
docker compose up -d
Key discovery: web = Desktop = mobile
When a user configures a custom connector on claude.ai (web), it automatically becomes available on Claude Desktop and Claude mobile — no separate configuration needed. The connector is linked to the user's Anthropic account, not to a specific app.
This means: one setup on the web covers all platforms.
Security
- Tokens are read from environment variables, never hardcoded
- Filter injection prevented by escaping single quotes
- Table names are URL-encoded to preserve accented characters
- API keys validated against TeamDesk table (not hardcoded)
- OAuth auto-approve uses PKCE S256 for security
- Rate limiting prevents abuse (configurable per-minute limit)
- See SECURITY.md for responsible disclosure
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.