Hi TeamDesk team,
I'd like to share two open-source projects I've built for integrating TeamDesk with AI assistants (Claude, GPT, etc.), and propose that TeamDesk evaluates them as a starting point for official tools.
What I Built
1. TeamDesk MCP Server (Model Context Protocol) https://github.com/Danielbluz/teamdesk-mcp-v2
1. A Python server (FastMCP) that exposes 11 tools for AI assistants to interact with TeamDesk databases in real time:
- list_tables — List all tables in the database
- describe_table — Get column names, types, and metadata
- get_records — Query records with filters, sorting, and pagination
- get_record — Retrieve a single record by ID
- select_view — Query records from a specific view
- search_records — Full-text search using Contains() filters
- create_record / update_record / delete_record — Full CRUD
- upsert_records — Create or update based on match column
- gerar_documento — Generate documents from templates
It works in two modes: local (stdio, for Claude Code / Claude Desktop) and remote (SSE, for multi-user access). Authentication uses standard TeamDesk Bearer tokens — no extra credentials needed.
We've been using this in production since January 2026 with 4 users across two companies, processing thousands of records daily. It handles accented column names (Portuguese), batch operations, and all the typical TeamDesk API quirks.
2. TeamDesk Skill for Claude Code
https://github.com/Danielbluz/teamdesk-skill
2. A comprehensive knowledge base that teaches AI assistants how to work with TeamDesk effectively. It includes:
- Complete REST API reference — All methods (select, create, update, upsert, delete, describe, attachment), parameters, response formats, error codes, plus undocumented endpoints like exportview.aspx and gateway.aspx
- Formula reference — 80+ functions with full syntax, examples, and type rules
- 20 community patterns — Curated from a systematic scan of the 100 most recent forum posts here on CRMDesk. Techniques shared by Slava, Kirill, Scott Miller, Jorge Sola, Pierre, Calvin Peters, and others
— organized into categories: UI enhancements (QuickChart.io charts, progress bars), workflow patterns (status flip, webhook callback), data patterns (custom auto-numbering, API batching), document generation (CardText, dynamic checkboxes), and more.
When this skill is installed, any AI assistant can write correct TeamDesk formulas, build proper API calls, and apply community-proven patterns — without hallucinating syntax or inventing nonexistent features.
The Idea
I'd love for the TeamDesk team to evaluate these two projects and consider using them as a foundation for official TeamDesk MCP and AI tools. Specifically:
1. Official MCP Server — An MCP server maintained by TeamDesk would give users seamless AI integration out of the box. My implementation covers read/write operations on existing tables, but it's limited to what the current REST API allows.
2. Official AI Skill/Knowledge Base — A curated, always-up-to-date reference that AI assistants can use. This would dramatically reduce the "hallucination problem" where AI makes up TeamDesk features that don't exist.
3. Schema management via API/MCP (Feature Request) — This is the big one. Currently, the REST API only supports CRUD operations on records within existing tables. There is no way to programmatically:
- Create or modify tables
- Add, rename, or remove columns
- Set column types, constraints, or default values
- Create or modify views
- Manage relationships (references, summaries, lookups)
This means AI assistants can read and write data, but cannot help users design or evolve their database schema. Adding schema management endpoints (even basic ones like create table, add column, set column type) would unlock a completely new level of AI-assisted database development — where an AI could not only query your data but also help you build the database structure itself.
Why This Matters
MCP (Model Context Protocol) is becoming the standard for connecting AI assistants to external tools. Anthropic, OpenAI, Google, and others are adopting it. Having an official TeamDesk MCP server would position TeamDesk as AI-ready, which is increasingly important for users evaluating database platforms.
The community patterns in the skill also represent significant tribal knowledge that is currently scattered across forum posts. Packaging it as a structured reference benefits everyone.
Both projects are MIT-licensed — the TeamDesk team is welcome to fork, adapt, or use them however they see fit.
Happy to discuss or help with any of this.
Daniel Luz |  |
