Unichat MCP Server in Python

1# Unichat MCP Server in Python
2Also available in [TypeScript](https://github.com/amidabuddha/unichat-ts-mcp-server)
3--
4 <h4 align="center">
5  <a href="https://mseep.ai/app/amidabuddha-unichat-mcp-server">
6  <img src="https://mseep.net/pr/amidabuddha-unichat-mcp-server-badge.png" alt="MseeP.ai Security Assessment Badge" />
7  </a>
8 </h4>
9  <h4 align="center">
10  <a href="https://github.com/amidabuddha/unichat-mcp-server/blob/main/LICENSE.md">
11  <img src="https://img.shields.io/github/license/amidabuddha/unichat-mcp-server" alt="Released under the MIT license." />
12  </a>
13  <a href="https://archestra.ai/mcp-catalog/amidabuddha__unichat-mcp-server">
14    <img src="https://archestra.ai/mcp-catalog/api/badge/quality/amidabuddha/unichat-mcp-server" alt="Trust Score" />
15  </a>
16  <a href="https://smithery.ai/server/unichat-mcp-server">
17    <img src="https://smithery.ai/badge/unichat-mcp-server" alt="Smithery Server Installations" />
18  </a>
19</h4>
20 <h4 align="center">
21  <a href="https://mcphub.com/mcp-servers/amidabuddha/unichat-mcp-server">
22  <img src="https://img.mcphub.com/_next/image?url=%2Flogo-dark.png&w=48&q=75" alt="Hosted at MCPHub" />
23  </a>
24 </h4>
25 
26Send requests to OpenAI, MistralAI, Anthropic, xAI, Google AI, DeepSeek, Alibaba, Inception using MCP protocol via tool or predefined prompts.
27Vendor API key required
28 
29### Tools
30 
31The server implements one tool:
32- `unichat`: Send a request to unichat
33  - Takes "messages" as required string arguments
34  - Returns a response
35 
36### Prompts
37 
38- `code_review`
39  - Review code for best practices, potential issues, and improvements
40  - Arguments:
41    - `code` (string, required): The code to review"
42- `document_code`
43  - Generate documentation for code including docstrings and comments
44  - Arguments:
45    - `code` (string, required): The code to comment"
46- `explain_code`
47  - Explain how a piece of code works in detail
48  - Arguments:
49    - `code` (string, required): The code to explain"
50- `code_rework`
51  - Apply requested changes to the provided code
52  - Arguments:
53    - `changes` (string, optional): The changes to apply"
54    - `code` (string, required): The code to rework"
55 
56## Quickstart
57 
58### Install
59 
60#### Claude Desktop
61 
62On MacOS: `~/Library/Application\ Support/Claude/claude_desktop_config.json`
63On Windows: `%APPDATA%/Claude/claude_desktop_config.json`
64 
65**Supported Models:**
66> A list of currently supported models to be used as `"SELECTED_UNICHAT_MODEL"` may be found [here](https://github.com/amidabuddha/unichat/blob/main/unichat/models.py). Please make sure to add the relevant vendor API key as `"YOUR_UNICHAT_API_KEY"`
67 
68**Example:**
69```json
70"env": {
71  "UNICHAT_MODEL": "gpt-4o-mini",
72  "UNICHAT_API_KEY": "YOUR_OPENAI_API_KEY"
73}
74```
75Development/Unpublished Servers Configuration
76```json
77"mcpServers": {
78  "unichat-mcp-server": {
79    "command": "uv",
80    "args": [
81      "--directory",
82      "{{your source code local directory}}/unichat-mcp-server",
83      "run",
84      "unichat-mcp-server"
85    ],
86    "env": {
87      "UNICHAT_MODEL": "SELECTED_UNICHAT_MODEL",
88      "UNICHAT_API_KEY": "YOUR_UNICHAT_API_KEY"
89    }
90  }
91}
92```
93 
94Published Servers Configuration
95```json
96"mcpServers": {
97  "unichat-mcp-server": {
98    "command": "uvx",
99    "args": [
100      "unichat-mcp-server"
101    ],
102    "env": {
103      "UNICHAT_MODEL": "SELECTED_UNICHAT_MODEL",
104      "UNICHAT_API_KEY": "YOUR_UNICHAT_API_KEY"
105    }
106  }
107}
108```
109 
110### Installing via Smithery
111 
112To install Unichat for Claude Desktop automatically via [Smithery](https://smithery.ai/server/unichat-mcp-server):
113 
114```bash
115npx -y @smithery/cli install unichat-mcp-server --client claude
116```
117 
118## Development
119 
120### Building and Publishing
121 
122To prepare the package for distribution:
123 
1241. Remove older builds:
125```bash
126rm -rf dist
127```
128 
1292. Sync dependencies and update lockfile:
130```bash
131uv sync
132```
133 
1343. Build package distributions:
135```bash
136uv build
137```
138 
139This will create source and wheel distributions in the `dist/` directory.
140 
1414. Publish to PyPI:
142```bash
143uv publish --token {{YOUR_PYPI_API_TOKEN}}
144```
145 
146### Debugging
147 
148Since MCP servers run over stdio, debugging can be challenging. For the best debugging
149experience, we strongly recommend using the [MCP Inspector](https://github.com/modelcontextprotocol/inspector).
150 
151 
152You can launch the MCP Inspector via [`npm`](https://docs.npmjs.com/downloading-and-installing-node-js-and-npm) with this command:
153 
154```bash
155npx @modelcontextprotocol/inspector uv --directory {{your source code local directory}}/unichat-mcp-server run unichat-mcp-server
156```
157 
158 
159Upon launching, the Inspector will display a URL that you can access in your browser to begin debugging.
160