log-reader-mcp
MCP server to start, monitor, search, and detect errors in logs from any project, even those without log files.
README
log-reader-mcp
MCP server for Claude Code — lets Claude start, monitor, search, and detect errors in logs from any project, even ones that don't write log files.
Requirements: Python 3.8+, Claude Code
Install (one-line, any machine)
SSH (if you have GitHub SSH keys set up):
git clone git@github.com:ankit-jhajhria/log-reader-mcp.git ~/.claude/mcp-servers/log-reader && bash ~/.claude/mcp-servers/log-reader/setup.sh
HTTPS (works without SSH keys):
git clone https://github.com/ankit-jhajhria/log-reader-mcp.git ~/.claude/mcp-servers/log-reader && bash ~/.claude/mcp-servers/log-reader/setup.sh
Restart Claude Code after install.
Ubuntu/Debian users: If setup fails with a
venverror, runsudo apt install python3-venvfirst — the script will tell you this too.
How to use in a new session
Project with no log files (Node, Go, Rust, plain Python, etc.)
Open Claude Code in your project and say:
run "node server.js" and capture the logs
Claude starts the server in the background, captures all terminal output, and tails the log immediately. Always set the working directory so the command runs from the right folder:
run "npm start" in /home/user/my-node-project and capture the logs
To check for problems:
detect errors in the captured log
To restart after a code change:
restart the server
To stop it:
stop the server
Project that already writes log files (Django, Rails, etc.)
tail logs/debug.log
detect errors in logs/debug.log
search logs/debug.log for "Traceback" with 3 lines of context
show me new lines in logs/debug.log since line 400
Don't know where the logs are
list log files in the logs folder
Files modified in the last 60 seconds are marked [active] so you can spot the right one instantly.
All tools
Process management
| Tool | What it does |
|---|---|
run_and_capture(command, cwd, log_file) |
Start any server in background, capture all output to a file |
list_captured_processes() |
Show all running/stopped captured processes |
stop_process(pid) |
Stop a process |
restart_process(pid) |
Stop and restart with the same command and working directory |
Log reading
| Tool | What it does |
|---|---|
detect_errors(path, last_lines) |
Auto-scan for ERROR, Exception, Traceback, WARNING, CRITICAL, fatal |
tail_file(path, lines=80) |
Last N lines of any log file |
get_new_lines(path, after_line) |
New lines since line N — use to poll a running server |
search_log(path, pattern, context_lines) |
Regex/string search with optional surrounding context lines |
list_log_files(directory, pattern) |
List log files with sizes and active status |
file_info(path) |
Size, modified time, line count |
Troubleshooting
python3-venv error on Ubuntu/Debian:
sudo apt install python3-venv
Permission denied reading a log file:
Some system logs (e.g. /var/log/syslog) require elevated permissions. Run Claude Code with sudo or copy the log to a readable location.
Process exits immediately after run_and_capture:
The tool detects this and shows you the output. Common causes: wrong command, port already in use, missing dependencies. Fix the command and try again.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。