Connhex MCP Server
Official# Connhex Tools
Tools workspace for [Connhex](https://connhex.com). Contains the MCP server,
the CLI, and the API SDK as sibling packages in a single
[uv workspace](https://docs.astral.sh/uv/concepts/workspaces/).
## Packages
| Package | Description | README |
| -------------------------------------- | ------------------------------------------------------------------------------- | ----------------------------------- |
| [`connhex-sdk`](packages/connhex-sdk/) | Async Python client for the Connhex API. Used internally by the other packages. | [→](packages/connhex-sdk/README.md) |
| [`connhex-mcp`](packages/connhex-mcp/) | MCP server that exposes Connhex APIs as tools for AI clients (Claude, etc.). | [→](packages/connhex-mcp/README.md) |
| [`connhex-cli`](packages/connhex-cli/) | Command-line interface for managing connhex resources. | [→](packages/connhex-cli/README.md) |
## Getting started
**Prerequisites:** Python 3.11+, [uv](https://docs.astral.sh/uv/getting-started/installation/).
```bash
git clone https://github.com/compiuta-origin/connhex-tools.git
cd connhex-tools
uv sync
```
This installs all three packages and their dependencies into a shared `.venv`.
## Development
```bash
uv run pytest # run all tests
uv run ruff check . && uv run ruff format .
uv run pre-commit run --all-files
```
Run a specific package's tests:
```bash
uv run pytest packages/connhex-cli/tests
```
TDQS
Scored across 30 tools
There is significant overlap between generic resource tools (create_resource, delete_resource, etc.) and manufacturing resource tools (create_manufacturing_resource, etc.), with identical descriptions. This creates ambiguity about when to use each set. Other tools like read_channel_messages and read_thing_messages are more distinct.
Nearly all tools follow a consistent verb_noun pattern (e.g., create_resource, list_things, get_rule). The only slight outlier is whoami, but it is a common convention. Naming is highly predictable and uniform across domains.
With 30 tools, the server covers multiple domains (generic resources, manufacturing, rules, things, models, messages, user info). However, the explicit duplication of CRUD tools for generic and manufacturing resources inflates the count unnecessarily. The scope is broad but the count is borderline high.
The tool surface has notable gaps: things can only be read/fetched (no create, update, or delete), models lack CRUD beyond get/list, and channels only support reading messages. Full CRUD is only present for resources (generic and manufacturing) and rules. The absence of basic lifecycle operations for core entities (things, models) will likely cause agent failures.