mcp-wireshark
Enables AI assistants to analyze, filter, and capture network traffic using Wireshark/tshark, allowing natural language interaction with packet captures.
README
mcp-wireshark
Community-maintained MCP server for Wireshark /
tshark. Not affiliated with Wireshark or Anthropic. Give your AI assistant direct access to packet captures. Ask Claude to summarize a.pcap, follow a TCP stream, filter for a specific protocol, or capture live traffic — all without leaving the chat.

Quick start with Claude Code
pip install mcp-wireshark
claude mcp add --transport stdio --scope user mcp-wireshark -- mcp-wireshark
That's it. Open Claude Code and try:
"Summarize ./capture.pcap and tell me which IPs talked the most."
--scope user makes the server available across every Claude Code project. Drop the flag to install it for the current project only. See claude mcp docs for more.
Verify the install
claude mcp list
You should see mcp-wireshark listed. Inside Claude Code, ask:
"Run check_installation."
If tshark is on your PATH, it returns the version. If not, see troubleshooting.
Tools
The server exposes 13 tools, split cleanly between read tools (safe, no side effects) and write tools (capture traffic or write files). Both groups are annotated with the standard MCP readOnlyHint so any compliant client can surface the distinction.
Read tools
Safe to call freely — they only inspect state.
| Tool | What it does |
|---|---|
check_installation |
Verify tshark is installed and show version |
list_interfaces |
List network interfaces available to capture from |
read_pcap |
Read packets from a .pcap / .pcapng file (preview + total count) |
display_filter |
Apply a Wireshark display filter to a pcap |
summarize_pcap |
High-level summary: I/O stats, protocol hierarchy, top talkers |
stats_by_proto |
Protocol hierarchy statistics |
follow_tcp |
Reassemble a TCP stream and return its payload |
follow_udp |
Reassemble a UDP stream and return its payload |
expert_info |
tshark expert analysis: warnings, errors, and notes grouped by severity |
decode_protocol |
Extract protocol fields as a TSV table. Curated defaults for HTTP, DNS, TLS, GOOSE, MMS, SV, SIP, ICMP; arbitrary fields for any other protocol |
protocol_stats |
Aggregate -z reports (protocol hierarchy, conversations, endpoints, HTTP/DNS/SMB stats) |
Write tools
These create files or capture live traffic. Compliant clients may prompt before invoking.
| Tool | What it does |
|---|---|
live_capture |
Capture live traffic from an interface (capped at 5 minutes / 10k packets) |
export_json |
Export packets from a pcap to a JSON file at a path you choose |
See it in action
These clips run the real tools against demo/demo.pcapng — a
short home-network capture. Regenerate them with python demo/render_gif.py <scene>.
summarize_pcap — characterize an unknown capture at a glance

decode_protocol — filter to a protocol and get a compact table (here: TLS SNI and DNS-over-HTTPS lookups)

expert_info — let tshark surface the warnings and anomalies for you

Example prompts
Drop these into Claude Code as-is:
List my network interfaces.
Summarize ./traffic.pcap.
From ./traffic.pcap, show me only HTTP requests.
Follow TCP stream 0 in ./traffic.pcap and tell me what protocol is in it.
Capture 30 seconds of traffic on Wi-Fi filtered to tcp.port == 443.
Export every DNS packet from ./traffic.pcap to ./dns.json.
Decode the GOOSE messages in ./substation.pcapng — only stNum >= 1.
Run expert analysis on ./traffic.pcap and group findings by severity.
Show me the IP conversations in ./traffic.pcap.
Useful display filters
| Filter | Matches |
|---|---|
tcp.port == 80 |
HTTP |
tcp.port == 443 |
HTTPS |
dns |
All DNS |
http.request |
HTTP requests only |
ip.addr == 10.0.0.1 |
Traffic to/from a specific host |
tcp.flags.syn == 1 && tcp.flags.ack == 0 |
TCP SYN packets only |
For substation engineers analyzing IEC 61850 traffic:
| Filter | Matches |
|---|---|
goose |
All GOOSE messages |
goose.stNum > 0 |
GOOSE messages with state changes |
mms |
All MMS traffic |
sv |
Sampled Values |
Other clients
Anything that speaks MCP works. The package installs an mcp-wireshark binary on PATH.
<details> <summary><b>Claude Desktop</b></summary>
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"wireshark": {
"command": "mcp-wireshark"
}
}
}
</details>
<details> <summary><b>VS Code (Copilot / GitHub Copilot Chat)</b></summary>
Create .vscode/mcp.json in your workspace:
{
"servers": {
"wireshark": {
"command": "mcp-wireshark"
}
}
}
</details>
<details> <summary><b>Cursor / Windsurf / others</b></summary>
Use the same stdio invocation: command: mcp-wireshark. No transport flags.
</details>
Prerequisites
- Python 3.10+
- Wireshark installed;
tsharkreachable onPATH
Install with pip or uv:
pip install mcp-wireshark
# or
uvx mcp-wireshark
Troubleshooting
<details> <summary><b>tshark not found on Windows</b></summary>
Add Wireshark to your system PATH:
- Press
Win+R→ runsysdm.cpl→ Advanced → Environment Variables - Edit
Path→ addC:\Program Files\Wireshark - Restart your terminal and Claude Code, then re-run
check_installation
(Avoid passing PATH through claude mcp add --env — values are taken literally, no %PATH% expansion.)
</details>
<details> <summary><b>Permission denied capturing on Linux</b></summary>
Add yourself to the wireshark group, then log out and back in:
sudo usermod -aG wireshark $USER
</details>
<details> <summary><b>"No packets captured" from live_capture</b></summary>
- Confirm the interface name from
list_interfaces(Wireshark uses different names thanifconfig/ip) - On macOS, you may need to install ChmodBPF (ships with the Wireshark
.dmg) - Check that no display filter is excluding everything </details>
Development
git clone https://github.com/khuynh22/mcp-wireshark.git
cd mcp-wireshark
python -m venv venv && source venv/bin/activate # Windows: venv\Scripts\activate
pip install -e ".[dev]"
pytest # tests
black src tests # format
ruff check src tests # lint
mypy src # type check
The codebase is organized so new tools land in one of two clearly-scoped files:
src/mcp_wireshark/read_tools.py— anything that just inspects statesrc/mcp_wireshark/write_tools.py— anything that captures traffic or writes files
server.py only contains routing. See CLAUDE.md and CONTRIBUTING.md.
Security
Every file path is validated (.. rejected, extension allow-listed). Every display filter is checked for shell metacharacters. tshark is always invoked via asyncio.create_subprocess_exec, never shell=True. Hard caps: 10k packets per call, 5 min per live capture. See SECURITY.md.
License
MIT — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.