biji-mcp
# Get笔记 MCP Server
在 Claude 中搜索 Get笔记 知识库内容。
## 安装
```bash
# 克隆项目
git clone <repo>
cd biji-mcp
# 安装依赖
uv sync
```
## 配置
首次运行会自动创建配置文件 `~/.biji-mcp/config.json`，请编辑填入你的API配置：
```json
{
"knowledge_bases": {
"工作": {
"token": "your-api-token",
"topic_id": "your-topic-id",
"description": "工作相关笔记"
}
},
"default": "工作",
"settings": {
"default_top_k": 10,
"timeout": 30
}
}
```
获取API配置：进入 Get笔记 Web版知识库 → 点击"API设置"按钮。
## Claude Desktop 配置
在 `~/Library/Application Support/Claude/claude_desktop_config.json` 添加：
```json
{
"mcpServers": {
"biji": {
"command": "uv",
"args": ["--directory", "/path/to/biji-mcp", "run", "biji-mcp"]
}
}
}
```
## 可用工具
### biji_search
在知识库中搜索，返回AI生成的答案。
参数：
- `question` (必需): 搜索问题
- `kb`: 知识库名称
- `deep_seek`: 启用深度思考
- `with_refs`: 返回引用来源
### biji_recall
召回原始内容片段，不经AI处理。
参数：
- `question` (必需): 搜索问题
- `kb`: 知识库名称
- `top_k`: 返回结果数量
- `intent_rewrite`: 意图改写
### biji_list_kb
列出已配置的知识库。
## 调试
设置环境变量启用调试日志：
```bash
BIJI_MCP_DEBUG=1 uv run biji-mcp
```
## API限制
- QPS: 2
- 日调用: 5000次
- 公测期免费
TDQS
Scored across 3 tools
Each tool has a clearly distinct purpose: biji_search returns AI-generated answers with citations, biji_recall returns raw content snippets ranked by relevance, and biji_list_kb lists configured knowledge bases. There is no overlap between these operations, so an agent can confidently choose the right tool based on whether it needs processed or raw results.
All tool names follow the biji_ prefix and a simple verb pattern: search, recall, list_kb. This is consistent and predictable. The only minor deviation is that list_kb uses an underscore while the others are single verbs, but the overall convention is uniform.
With 3 tools, the server is well-scoped for a knowledge base search service. Each tool serves a distinct need—AI search, raw recall, and KB discovery—without unnecessary extras. This is within the ideal 3-15 range and feels complete for its niche purpose.
The server covers the core functionality of searching and recalling knowledge base content, plus listing available KBs. A minor gap is the lack of a tool to retrieve metadata or browse specific documents within a KB, but that is beyond the primary search/recall workflow and can be worked around.