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., "@noc-pcap-mcp-server Summarize the packet capture at /tmp/capture.pcap"
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.
An MCP (Model Context Protocol) server that lets an LLM-based assistant diagnose network problems from packet captures (PCAP/PCAPNG files), without the operator needing to use Wireshark directly. Built for junior analysts in a NOC (Network Operations Center) setting.
Built with the official Python MCP SDK
(FastMCP) and scapy for packet parsing.
All tools take a file_path to a .pcap/.pcapng file. stream_id values
are returned by list_tcp_conversations and identify one TCP conversation.
Tool | Parameters | Returns |
|
| Duration, packet count, protocol breakdown, top talkers |
|
| List of TCP conversations (4-tuple, |
|
| RTT, retransmission count, zero-window events, duplicate ACKs, with the frame numbers that evidence each finding |
|
| Findings ranked by severity: port scans, ARP spoofing, high-entropy DNS queries, cleartext credentials |
|
| Reassembled JSON-RPC messages, classified (initialize / notification / request / response) and paired by |
These are designed to be composable: a host LLM chains them (e.g. list conversations, then analyze the relevant one) rather than calling one do-everything tool.
Related MCP server: Wireshark-MCP
Python >= 3.11
uv (recommended) or pip
git clone https://github.com/anthonylouschwank/noc-pcap-mcp-server.git
cd noc-pcap-mcp-server
uv syncOr with pip:
python -m venv .venv
.venv\Scripts\activate # Windows
pip install -e .Run over stdio:
uv run noc-pcap-mcpExample Claude Desktop config entry:
{
"mcpServers": {
"noc-pcap": {
"command": "uv",
"args": ["run", "--directory", "/absolute/path/to/noc-pcap-mcp-server", "noc-pcap-mcp"]
}
}
}See noc-assistant,
which launches this server as a subprocess and aggregates its tools with
other MCP servers.
src/noc_pcap_mcp/
├── server.py # tool registration (FastMCP)
├── pcap_utils.py # capture summary + TCP conversation indexing
├── tcp_analysis.py # RTT / retransmissions / zero-window per conversation
├── security.py # port scans, ARP spoofing, DNS entropy, cleartext creds
└── jsonrpc.py # TCP reassembly + JSON-RPC message classificationScaffolding stage: tools are registered with their final signatures and
docstrings; the analysis logic in pcap_utils.py, tcp_analysis.py,
security.py and jsonrpc.py is not implemented yet.
Built by Anthony Schwank for CC3067 - Redes, Universidad del Valle de Guatemala.
RTT, retransmissions, zero-window events and duplicate ACKs for one TCP conversation, citing the frame numbers that evidence each finding.
Inputs: file_path (required), stream_id (required).
Scan a PCAP for port scans, ARP spoofing, high-entropy DNS queries and cleartext credentials, ranked by severity.
Inputs: file_path (required). Returns: result
Reassemble a TCP conversation's payload, classify JSON-RPC messages and pair requests with responses by id.
Inputs: file_path (required), stream_id (required).
Overview of a PCAP file: duration, packet count, protocols seen, top talkers.
Inputs: file_path (required).
Enumerate TCP conversations (4-tuple) in a PCAP with packet/byte counts and timing.
Inputs: file_path (required). Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 5 tools
Each tool targets a distinct layer or task: overall summary, conversation enumeration, per-conversation transport analysis, security scanning, and JSON-RPC payload extraction. Though analyze_tcp_conversation and extract_json_rpc_exchange both work on a conversation, one is transport-focused and the other application-focused, so no reasonable misselection exists.
All tool names follow a clear verb_noun snake_case pattern: get/list/analyze/detect/extract plus a specific object. The singular 'tcp_conversation' after 'analyze_' and the plural in 'list_tcp_conversations' is grammatically consistent, not a style break.
Five tools is a well-scoped set for a PCAP analysis server; each tool provides a distinct capability without redundancy. This is within the ideal range and leaves room for focused use.
The surface covers the core NOC/PCAP workflow: summarize, enumerate conversations, drill into a conversation's health, detect anomalies, and extract application-layer exchanges. A minor gap is the lack of generic packet/filter drill-down or pcap file management, but agents can work around these.
IP intelligence, network diagnostics, and AI service availability tools for AI agents.
Network, domain and website diagnostics for AI clients via MCP.
Provide detailed Pokémon data and information through a standardized MCP interface. Enable LLMs an…
Scan prompts for injection attacks, redact PII, and audit LLM SDK usage from any MCP client