Content
# UniFuncs MCP Server
MCP Server for the UniFuncs API - Enhanced with Deep Search and Deep Research capabilities
## Features
This MCP server provides access to the following UniFuncs APIs:
### 1. Web Search (`web-search`)
Real-time web search with comprehensive results
- Search across the internet with keywords
- Filter by freshness (Day/Week/Month/Year)
- Pagination support (1-50 results per page)
- Multiple output formats (JSON/Markdown/Text)
### 2. Web Reader (`web-reader`)
Extract detailed content from web pages
- Clean content extraction
- Optional image inclusion
- Link summary support
- Markdown output
### 3. Deep Search - Sync (`deep-search-sync`)
Real-time deep search with immediate results
- Model: S3 / S3 Pro
- Streaming support
- Instant response
### 4. Deep Search - Async (`deep-search-create-task` + `deep-search-query-task`)
Asynchronous deep search for complex queries
- Create task and get task_id immediately
- Poll for task status and results
- Background processing for large queries
### 5. Deep Research (`deep-research-create-task` + `deep-research-query-task`)
Comprehensive deep research capabilities
- Models: U3, U3 Pro
- Customizable research parameters:
- Introduction: Set researcher persona
- Reference style: link/number/footnote
- Max depth: Up to 50 iterations (25 recommended)
- Domain scope: Limit search to specific domains
- Domain blacklist: Exclude specific domains
- Custom output prompts
- Important URLs, keywords, and prompts
- Async task management
## Setup
### API Key
Get a UniFuncs API Key:
https://unifuncs.com/account
### NPX (STDIO)
```json
{
"mcpServers": {
"unifuncs": {
"command": "npx",
"args": [
"-y",
"@unifuncs/ufn-mcp-server"
],
"env": {
"UNIFUNCS_API_KEY": "sk-**********"
}
}
}
}
```
Note: the `npx` experience reflects the latest package published to npm, not necessarily the current `main` branch. If the published npm version is behind `main`, use the source build instructions below to reproduce the latest merged toolset.
### Source Build (Local)
```bash
git clone https://github.com/UniFuncs/ufn-mcp-server.git
cd ufn-mcp-server
yarn install
npm run build
UNIFUNCS_API_KEY=sk-********** node build/index.js
```
Example Codex MCP config:
```toml
[mcp_servers.unifuncs-local]
command = "node"
args = ["/absolute/path/to/ufn-mcp-server/build/index.js"]
[mcp_servers.unifuncs-local.env]
UNIFUNCS_API_KEY = "sk-**********"
```
### SSE Server
For SSE transport, set the environment variable:
```bash
export UNIFUNCS_SSE_SERVER=true
export UNIFUNCS_SSE_SERVER_PORT=5656 # Optional, default is 5656
```
Or use the `--sse` flag:
```bash
npx @unifuncs/ufn-mcp-server --sse
```
## Tool Reference
### web-search
```
Query: Search keywords
Freshness: Day | Week | Month | Year (optional)
Page: Page number, default 1 (optional)
Count: Results per page, 1-50, default 10 (optional)
Format: json | markdown | text, default json (optional)
```
### web-reader
```
URL: Page URL to read
Format: markdown (optional)
IncludeImages: boolean (optional)
LinkSummary: boolean (optional)
```
### deep-search-sync
```
Model: s3 / s3-pro (default: s3)
Messages: Array of {role: "user"|"assistant"|"system", content: string}
Stream: boolean (default: false)
```
### deep-search-create-task
```
Model: s3 / s3-pro (default: s3)
Messages: Array of {role: "user"|"assistant"|"system", content: string}
Returns: task_id for querying status
```
### deep-search-query-task
```
Task_ID: Task ID from create_task
Returns: Task status, progress, and results when completed
```
### deep-research-create-task
```
Model: u3 | u3-pro (default: u3)
Content: Research question/topic
Introduction: Researcher persona (optional)
Reference_Style: link | number | footnote (default: link)
Generate_Summary: boolean (default: false)
Max_Depth: 1-50 (default: 25, recommended)
Domain_Scope: Comma-separated domains (optional)
Domain_Blacklist: Comma-separated domains to exclude (optional)
Output_Prompt: Custom output template (optional)
Important_URLs: Comma-separated URLs (optional)
Important_Keywords: Comma-separated keywords (optional)
Important_Prompt: Important prompt content (optional)
Push_To_Share: boolean (default: false)
Set_Public: boolean (default: false)
Returns: task_id for querying status
```
### deep-research-query-task
```
Task_ID: Task ID from create_task
Returns: Task status, progress, and results when completed
```
## Examples
### Web Search
```javascript
{
"query": "OpenClaw AI",
"count": 5,
"format": "json"
}
```
### Deep Search Async
```javascript
// Create task
{
"model": "s3",
"messages": [
{ "role": "user", "content": "What are the latest developments in AI?" }
]
}
// Query task (use returned task_id)
{
"task_id": "3aff2a91-7795-4b73-8dab-0593551a27a1"
}
```
### Deep Research
```javascript
// Create research task
{
"model": "u3",
"content": "Analyze the impact of AI on healthcare",
"max_depth": 25,
"domain_scope": "arxiv.org, nature.com",
"generate_summary": true
}
// Query research task
{
"task_id": "research-task-id-here"
}
```
## Pricing
- Web Search: Pay per request
- Web Reader: Pay per request
- Deep Search: Pay per token usage
- Deep Research:
- U3: M Tokens
- U3 Pro: M Tokens
For detailed pricing, visit: https://unifuncs.com/pricing
## Support
- Documentation: https://unifuncs.com/api
- GitHub: https://github.com/UniFuncs/ufn-mcp-server
- Email: service@unifuncs.com
- WeChat: unifuncs
## Maintainer Release Checklist
Before documenting new tools in the `npx` setup section, make sure the npm package has been published and smoke-tested.
1. Bump the package version.
2. Install dependencies and run `npm run build`.
3. Publish the package to npm.
4. Verify `npm view @unifuncs/ufn-mcp-server version` matches the intended release.
5. Verify `npx -y @unifuncs/ufn-mcp-server` exposes the same tools described in this README.
6. Update the changelog and README after the npm release is available.
## License
MIT
## Changelog
### v0.1.0 (2026-02-25)
- Added Deep Search sync API support
- Added Deep Search async API (create_task + query_task)
- Added Deep Research async API (create_task + query_task)
- Enhanced documentation with all tool references
- Improved error handling
### v0.0.6
- Initial release with web-search and web-reader
Connection Info
You Might Also Like
everything-claude-code
Complete Claude Code configuration collection - agents, skills, hooks,...
markitdown
MarkItDown-MCP is a lightweight server for converting URIs to Markdown.
cc-switch
All-in-One Assistant for Claude Code, Codex & Gemini CLI across platforms.
servers
Model Context Protocol Servers
servers
Model Context Protocol Servers
Time
A Model Context Protocol server for time and timezone conversions.