This server is a general-purpose MCP toolkit that lets an AI assistant control a computer, automate a browser, process data, run commands, and access the web.
Browser automation: open/close sessions, navigate, click, type, select, fill forms, take screenshots, run JavaScript, read console logs, manage cookies/tabs, monitor network requests, and inspect page content.
Computer use: capture screens, control mouse/keyboard, manage windows, and use the clipboard (via the Computer Use category).
File system: read, write, append, search, copy, delete, compare files, and manage directories.
Data processing: parse/convert/format JSON, CSV, XML, YAML, TOML; query and merge JSON; validate schemas.
Text processing: regex, encoding, email/URL extraction, text similarity, and diffing text.
Web & network: web search (with fallback engines and caching), fetch pages, extract titles/links, download files, check URL status, get headers, and make custom HTTP requests.
System monitoring: get system/CPU/memory/disk/process info and manage environment variables.
Utilities: generate UUIDs, hashes, random strings, passwords; date/time conversions; safe math evaluation.
Subagent AI orchestration: call external AI models (OpenAI/Anthropic), run parallel subtasks, and do conditional branching.
Command execution: run allowlisted commands securely with sanitization, timeouts, and audit logging.
Compression: create/extract ZIP/TAR archives with security features.
Provides web search capabilities using DuckDuckGo's search engine.
Enables delegation of subtasks to OpenAI's AI models for parallel execution and conditional branching.
Provides browser automation capabilities including page navigation, element interaction, screenshots, JavaScript execution, and multi-tab management using Selenium WebDriver.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@oh-my-mcp search the web for latest AI breakthroughs"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
English | ไธญๆ
A powerful Model Context Protocol (MCP) server with 154 practical tools across 11 categories, built using FastMCP.
oh-my-mcp provides tools for:
๐ Browser Automation (33 tools): Selenium-based browser automation: navigation, interaction, screenshots, JS execution, console logs, cookies, network monitoring, form filling, multi-tab management
๐ฆ Compression (5 tools): ZIP/TAR compression and extraction with security features
๐ฅ๏ธ Computer Use (26 tools): AI desktop control: screen capture (multi-monitor, base64/file), mouse control (move/click/drag/scroll), keyboard input (typing/keys/hotkeys), clipboard access, window management, safety failsafe configuration
๐ Data Processing (15 tools): JSON, CSV, XML, YAML, TOML parsing and manipulation
โก Command Execution (5 tools): Secure allowlist-based command execution: run whitelisted commands with argument sanitization, timeout protection, output size limits and audit logging; manage the persistent command allowlist
๐ File System (16 tools): Read, write, search files and directories, file comparison
๐ค Subagent AI Orchestration (6 tools): Delegate subtasks to external AI models with parallel execution and cost tracking
๐ป System (11 tools): System info, CPU/memory monitoring, environment variables
๐ Text Processing (9 tools): Regex, encoding, email/URL extraction, text similarity
๐ ๏ธ Utilities (10 tools): UUID, hashing, date/time operations, math, password generation
๐ Web & Network (18 tools): Web search, page fetching, HTML parsing, downloads, HTTP API client, DNS lookup
Related MCP server: mcp-web-calc
The full index lives in docs/README.md โ every topic is
available in English (X.md, default) and Chinese (X.zh.md). Highlights:
Document | |
Installation | |
Tool Reference (all 146 tools) | |
Computer Use Guide | |
Setup Guide (Claude Desktop) | |
Command Execution | CONFIGURATION.md ยท see Tool Reference |
Build Guide | |
Architecture | |
Subagent Guide | |
Contributing | |
Changelog |
Tool counts and descriptions in the docs are generated from code โ see Documentation Generation.
git clone https://github.com/quyansiyuanwang/oh-my-mcp.git
cd oh-my-mcp
pip install -e . # or: uv sync --all-extrasConfigure Claude Desktop:
python -m mcp_server.cli.config --claudeStart the server:
python -m mcp_server.mainDetails: Installation Guide and Setup Guide.
Logs are configured in mcp_server/utils.py โ level, destinations and format.
Path validation: prevents path traversal attacks
Safe evaluation: math expressions only allow whitelisted operations
Masked values: sensitive environment variables are masked
Confirmation required: file deletion requires confirm=True
Allowlisted execution: commands only run after explicit trust
Retry logic: network operations retry up to 3 times
All tools return JSON with descriptive messages instead of raising:
ValidationError โ invalid input parameters
NetworkError โ network request failures
FileOperationError โ file system errors
DataProcessingError โ data parsing/conversion errors
oh-my-mcp/
โโโ pyproject.toml # Dependencies & tool config
โโโ configure.py # Interactive setup wizard
โโโ src/
โโโ mcp_server/
โโโ main.py # Server entry point
โโโ utils.py # Infrastructure & utilities
โโโ command_executor.py # Secure command execution
โโโ cli/config.py # Configuration generator
โโโ tools/ # Tool plugins (11 categories, auto-discovered)Full layout: PROJECT_STRUCTURE.md.
from mcp_server.tools.registry import tool_handler
@tool_handler
def your_tool(param: str) -> str:
"""Tool description.
Args:
param: Parameter description
Returns:
Return value description
"""
...Tool counts and descriptions in the docs are generated from code (AST
analysis of @tool_handler docstrings + each plugin's config.yaml). After
adding, removing, or renaming tools run:
python scripts/docs/generate_docs.py --writeCI verifies docs are fresh with --check and fails if they are out of date.
Generated fragments live between <!-- DOCGEN:...:start/end --> markers; do
not edit inside them.
See CONTRIBUTING.md. Client configuration details: SETUP_GUIDE.md.
This project is provided as-is for educational and practical use.
Enjoy oh-my-mcp! ๐
Append content to a file.
Inputs: path (required), content (required), encoding. Returns: result
Navigate back in browser history.
Inputs: session_id (required). Returns: result
Click an element on the page.
Waits for the element to be clickable before clicking.
Inputs: session_id (required), selector (required), by, timeout. Returns: result
Close a browser session.
Closes the browser window and cleans up all associated resources.
Inputs: session_id (required). Returns: result
Close a browser tab.
Inputs: session_id (required), tab_index. Returns: result
่ทๅๆต่งๅจ้ ็ฝฎ่ฎพ็ฝฎ
Inputs: key. Returns: result
้็ฝฎๆต่งๅจ้ ็ฝฎไธบ้ป่ฎคๅผ
Returns: JSONๆ ผๅผ็ๆไฝ็ปๆ
Inputs: no named parameters. Returns: result
่ฎพ็ฝฎๆต่งๅจ้ ็ฝฎ๏ผไฟๅญๅฐ้ ็ฝฎๆไปถ๏ผ
Inputs: key (required), value (required). Returns: result
Delete browser cookies.
Inputs: session_id (required), name. Returns: result
Enable network request logging.
After enabling, network requests will be captured and can be retrieved using browser_get_network_logs.
Inputs: session_id (required). Returns: result
Execute JavaScript code in the page context.
Inputs: session_id (required), script (required), args. Returns: result
Fill multiple form fields at once.
Inputs: session_id (required), form_data (required), by. Returns: result
Find multiple elements matching a selector.
Inputs: session_id (required), selector (required), by, limit. Returns: result
Navigate forward in browser history.
Inputs: session_id (required). Returns: result
Get browser console logs.
Captures console.log, console.error, console.warn, etc. from the page.
Inputs: session_id (required), level. Returns: result
Get browser cookies.
Inputs: session_id (required), name. Returns: result
Get an attribute value from an element.
Inputs: session_id (required), selector (required), attribute (required), by. Returns: result
Get captured network request logs.
Requires browser_enable_network_log to be called first.
Inputs: session_id (required), filter_url, filter_method, limit. Returns: result
Get the HTML source code of the current page.
Inputs: session_id (required), max_length. Returns: result
Get the text content of an element.
Inputs: session_id (required), selector, by, max_length. Returns: result
Get the current page URL and title.
Inputs: session_id (required). Returns: result
List all active browser sessions.
Returns information about all currently active browser sessions including their URLs, titles, and configuration.
Returns: JSON string with list of active sessions
Inputs: no named parameters. Returns: result
List all open tabs in the browser session.
Inputs: session_id (required). Returns: result
Open a new browser tab.
Inputs: session_id (required), url. Returns: result
Open a new browser session and navigate to a URL.
Creates a new browser instance (Chrome or Edge) and navigates to the specified URL. Returns a unique session ID for subsequent operations.
Inputs: url (required), browser, headless, window_size, +2 more. Returns: result
Refresh the current page.
Inputs: session_id (required). Returns: result
Take a screenshot of the page or a specific element.
Inputs: session_id (required), save_path, selector, by, +2 more. Returns: result
Scroll the page or scroll an element into view.
Inputs: session_id (required), direction, amount, selector, +1 more. Returns: result
Select an option from a dropdown/select element.
Inputs: session_id (required), selector (required), value (required), by, +2 more. Returns: result
Set a browser cookie.
Inputs: session_id (required), name (required), value (required), domain, +4 more. Returns: result
Switch to a different browser tab by index.
Inputs: session_id (required), tab_index (required). Returns: result
Type text into an input element.
Inputs: session_id (required), selector (required), text (required), by, +2 more. Returns: result
Wait for an element to satisfy a condition.
Inputs: session_id (required), selector (required), by, timeout, +1 more. Returns: result
Calculate difference between two dates.
Inputs: date1 (required), date2 (required), unit. Returns: result
Safely evaluate a mathematical expression.
Inputs: expression (required). Returns: result
Calculate similarity between two text strings.
Inputs: text1 (required), text2 (required), method. Returns: result
Check password strength and provide recommendations.
Inputs: password (required). Returns: result
Check the HTTP status of a URL.
Inputs: url (required), timeout. Returns: result
Clear the search cache.
This will remove all cached search results, forcing fresh searches.
Returns: JSON string with operation result
Inputs: no named parameters. Returns: result
Create a TAR archive from files.
Inputs: files (required), output_path (required), compression. Returns: result
Create a ZIP archive from files.
Inputs: files (required), output_path (required), compression_level. Returns: result
Copy a file from source to destination.
Inputs: source (required), destination (required), overwrite. Returns: result
Count words and provide text statistics.
Inputs: text (required), detailed. Returns: result
Create a directory.
Inputs: path (required), parents. Returns: result
Convert CSV data to JSON.
Inputs: csv_string (required), delimiter, has_header. Returns: result
Convert date string to Unix timestamp.
Inputs: date_string (required), timezone. Returns: result
Decode Base64 to text.
Inputs: encoded (required), encoding. Returns: result
Delete a file (requires confirmation).
Inputs: path (required), confirm. Returns: result
Compare two files and show differences.
Inputs: file1 (required), file2 (required), context_lines, format. Returns: result
Compare two text strings and show differences.
Inputs: text1 (required), text2 (required), format. Returns: result
Perform DNS lookup.
Inputs: hostname (required), record_type. Returns: result
Download a file from URL and save to disk.
Inputs: url (required), save_path (required), timeout. Returns: result
Encode text to Base64.
Inputs: text (required), encoding. Returns: result
Extract all email addresses from text.
Inputs: text (required). Returns: result
Extract a TAR archive.
Inputs: tar_path (required), extract_to. Returns: result
Extract all URLs from text.
Inputs: text (required). Returns: result
Extract a ZIP archive.
Inputs: zip_path (required), extract_to, password. Returns: result
Fetch the HTML content of a webpage.
Inputs: url (required), timeout. Returns: result
Fetch and extract clean text content from a webpage.
Inputs: url (required), timeout. Returns: result
Check if a file or directory exists.
Inputs: path (required). Returns: result
Flatten nested JSON object into single-level object.
Inputs: json_string (required), separator. Returns: result
Format a date string using custom format.
Inputs: date_string (required), format. Returns: result
Format JSON with pretty printing.
Inputs: json_string (required), indent, sort_keys. Returns: result
Generate hash of text using specified algorithm.
Inputs: text (required), algorithm, encoding. Returns: result
Generate a strong password.
Inputs: length, include_symbols, include_numbers, exclude_ambiguous. Returns: result
Generate a random string.
Inputs: length, charset. Returns: result
Generate a UUID.
Inputs: version, uppercase. Returns: result
Get CPU information and current usage.
Returns: JSON string containing CPU details and usage
Inputs: no named parameters. Returns: result
Get current date and time.
Inputs: timezone, format. Returns: result
Get disk space information for a path.
Inputs: path. Returns: result
Get an environment variable value.
Inputs: name (required), default. Returns: result
Get detailed information about a file or directory.
Inputs: path (required). Returns: result
Get HTTP headers from a URL.
Inputs: url (required), timeout. Returns: result
Get memory (RAM) usage statistics.
Returns: JSON string containing memory usage details
Inputs: no named parameters. Returns: result
Get network interface information.
Returns: JSON string with network interfaces, IP addresses, and MAC addresses
Inputs: no named parameters. Returns: result
Extract all links from a webpage.
Inputs: url (required), timeout, absolute. Returns: result
Extract the title from a webpage.
Inputs: url (required), timeout. Returns: result
Get information about the current process.
Returns: JSON string with process details
Inputs: no named parameters. Returns: result
Get search cache and rate limiter statistics.
Returns: JSON string with cache stats, hit rate, and rate limiter info
Inputs: no named parameters. Returns: result
Get comprehensive system information.
Returns: JSON string containing system details
Inputs: no named parameters. Returns: result
Make HTTP request with custom headers and body.
Inputs: url (required), method, headers, body, +1 more. Returns: result
Extract value from JSON using dot notation path.
Inputs: json_string (required), path (required). Returns: result
Convert JSON array to CSV format.
Inputs: json_string (required). Returns: result
Convert JSON to YAML format.
Inputs: json_string (required). Returns: result
List contents of an archive file without extracting.
Inputs: archive_path (required). Returns: result
List contents of a directory.
Inputs: path, pattern, recursive. Returns: result
List all environment variables.
Inputs: filter_pattern. Returns: result
Merge two JSON objects.
Inputs: json_string1 (required), json_string2 (required), deep. Returns: result
Parse CSV data and return as JSON.
Inputs: csv_string (required), delimiter. Returns: result
Parse HTML and extract elements using CSS selector.
Inputs: html (required), selector (required). Returns: result
Parse and validate a JSON string.
Inputs: json_string (required). Returns: result
Parse TOML configuration file to JSON.
Inputs: toml_string (required). Returns: result
Parse a URL and extract its components.
Inputs: url (required). Returns: result
Parse YAML string to JSON.
Inputs: yaml_string (required). Returns: result
Read the contents of a file.
Inputs: path (required), encoding. Returns: result
Find all matches of a regular expression in text.
Inputs: text (required), pattern (required), flags. Returns: result
Replace text matching a regular expression.
Inputs: text (required), pattern (required), replacement (required), flags. Returns: result
Search for files in a directory.
Inputs: directory, pattern, name_contains. Returns: result
Call an external AI model to handle a subtask.
Supports OpenAI and Anthropic APIs with custom endpoint configuration. Returns response with token usage statistics.
Inputs: provider (required), model (required), messages (required), max_tokens, +1 more. Returns: result
Execute conditional branching based on AI decision.
First calls AI to evaluate a condition, then executes either true_task or false_task based on the decision. Useful for dynamic workflow control.
Inputs: condition_task (required), true_task (required), false_task (required). Returns: result
่ทๅๆๅฎๆไพๅ็ API ้ ็ฝฎไฟกๆฏ
Inputs: provider (required). Returns: result
ๅๅบๆๆๅทฒ้ ็ฝฎ็ AI ๆไพๅ
่ฟๅๆๆ้ ็ฝฎ็ๆไพๅๅๅ ถ็ถๆ๏ผๅ ๆฌๅฏ้ฅ้ข่งๅ้ ็ฝฎๆฅๆบใ
Returns: JSON ๆ ผๅผ็ๆไพๅๅ่กจ
Example: providers = subagent_config_list()
Inputs: no named parameters. Returns: result
่ฎพ็ฝฎ Subagent ๆไพๅ็ API ้ ็ฝฎ๏ผๆไน ๅไฟๅญ๏ผ
ๆญคๅทฅๅ ทๅฐ API ๅฏ้ฅๅๅบ็ก URL ไฟๅญๅฐ้ ็ฝฎๆไปถไธญ๏ผไธๆฌกๅฏๅจๆถ่ชๅจๅ ่ฝฝใ ้ ็ฝฎๆไปถไฝ็ฝฎ๏ผ~/.subagent_config.json
Inputs: provider (required), api_key (required), api_base. Returns: result
Execute multiple AI subtasks in parallel with result aggregation.
Coordinates concurrent calls to different AI models and aggregates results. Useful for breaking complex problems into independent subtasks.
Inputs: tasks (required), max_workers. Returns: result
Summarize or truncate text to a maximum length.
Inputs: text (required), max_length, method. Returns: result
Convert Unix timestamp to readable date.
Inputs: timestamp (required), format, timezone. Returns: result
Convert TOML to formatted JSON.
Inputs: toml_string (required), indent. Returns: result
Validate JSON syntax and structure.
Inputs: json_string (required). Returns: result
Validate if a string is a properly formatted URL.
Inputs: url (required). Returns: result
Search the web using multiple search engines withๆบ่ฝ fallback.
Features:
Automatic fallback between search engines (DuckDuckGo -> Bing)
Intelligent caching to reduce API calls
Rate limiting protection
Result deduplication
Inputs: query (required), max_results. Returns: result
Advanced web search with multi-engine support and parallel searching.
Features:
Support for multiple search engines: DuckDuckGo, Bing, Google, Baidu
Parallel searching across multiple engines
Intelligent result merging and deduplication
Smart caching mechanism
Rate limiting protection
Inputs: query (required), max_results, engines, parallel, +1 more. Returns: result
Search for news articles using multiple search engines with fallback.
Features:
Automatic fallback between news search engines
Intelligent caching
Rate limiting protection
Result deduplication
Inputs: query (required), max_results. Returns: result
Write content to a file.
Inputs: path (required), content (required), encoding, overwrite. Returns: result
Convert XML to JSON format.
Inputs: xml_string (required). Returns: result
Convert YAML to formatted JSON.
Inputs: yaml_string (required), indent. Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelogโฆ
Scored across 116 tools
Most tools have distinct purposes, especially in the browser and subagent groups. However, there is some overlap between similar tools like fetch_webpage/fetch_webpage_text, parse_csv/csv_to_json, and multiple search tools (web_search, web_search_advanced, web_search_news). Descriptions help but some boundaries are blurry.
All tool names follow a consistent verb_noun pattern in snake_case (e.g., browser_navigate, parse_csv, get_system_info). No mixed conventions or irregular naming, making the set predictable for an agent.
With 116 tools, the server is extremely large for an MCP server. While it aims to be a comprehensive toolkit, the sheer number likely overwhelms agents and increases selection errors. A more focused scope or modularization would improve coherence.
The tool set covers a wide range of domains (browser, file, text, system, web search, subagent) with good depth. Minor gaps exist, such as lack of in-place file editing or image processing, but overall the surface is quite complete for its stated purpose.
8 MCP servers, 104+ tools: memory, social, PDF, email, images, calendar, scheduler, files.
AI tool directory MCP server, live since July 2026. Tools, agents, models, companies, services.
MCP registry & directory: search, find & install 31k+ MCP servers & tools. Catalog and marketplace.
Nifty's MCP server โ exposes tasks, projects, messages, and files as tools for AI agents.
