pcapport

pcapport

pcapport analyses PCAP capture files entirely on your own machine. Point it at a classic libpcap file and it summarises what is inside: DNS questions and response codes, DHCP message types and the servers and clients involved, ICMP traffic, TCP connections with state estimates, TCP anomalies and retransmissions, traffic flow to and from one server, SIP calls, and basic capture information such as packet count, duration, and encapsulation.

There is no network access, no API key, and no live capture: the tool only reads the capture file you name. The parser is an original minimal implementation covering Ethernet (including a single VLAN tag) and raw-IP captures carrying IPv4, with TCP, UDP, and ICMP decoded, and DNS, DHCP, and SIP payloads understood.

Install

npx @reneturcios/pcapport analyze_capinfos '{"pcap_file":"capture.pcap"}'

Or run the container image:

docker run -i ghcr.io/cloudcorprecords/pcapport

The container defaults to the MCP server. Pass CLI arguments to use it as a CLI instead.

CLI

The CLI mirrors the tools: one subcommand per tool, taking its arguments as a single JSON object.

pcapport analyze_capinfos '{"pcap_file":"capture.pcap"}'
pcapport analyze_dns_packets '{"pcap_file":"capture.pcap"}'
pcapport analyze_tcp_connections '{"pcap_file":"capture.pcap","detailed":true}'
pcapport analyze_tcp_retransmissions '{"pcap_file":"capture.pcap","threshold":0.05}'
pcapport analyze_traffic_flow '{"pcap_file":"capture.pcap","server_ip":"10.0.0.5","server_port":80}'
pcapport --version

Exit codes: 0 on success, 1 on an error (the error names the problem, e.g. a bad PCAP magic or a truncated file), 2 for usage errors. An empty capture is not an error: tools return zeroed summaries for it.

MCP server

One bin serves both faces. The mcp subcommand starts a stdio MCP server:

npx -y @reneturcios/pcapport mcp

MCP client configuration:

{
  "mcpServers": {
    "pcapport": {
      "command": "npx",
      "args": ["-y", "@reneturcios/pcapport", "mcp"]
    }
  }
}

Tools

Tool	Input	What it does
analyze_dns_packets	pcap_file (required)	DNS query/response counts, top questions (aggregated, top 20), and a response-code summary.
analyze_dhcp_packets	pcap_file (required)	DHCP packet count by message type (discover, offer, request, ack, ...), server IPs, and client MACs.
analyze_icmp_packets	pcap_file (required)	ICMP counts by type, echo requests/replies, unreachable count, and top talkers.
analyze_tcp_connections	pcap_file (required), server_ip, server_port, detailed	Connections keyed by the normalised 4-tuple with client/server, packets, bytes, SYN/FIN/RST counts, and a state estimate (established, closed, reset, half-open). With detailed, a per-connection timestamp summary is included.
analyze_tcp_anomalies	pcap_file (required), server_ip, server_port	Counts of reset connections, half-open connections (SYN with no SYN-ACK), connections without data, and zero-window advertisements, computed from flags and window fields.
analyze_tcp_retransmissions	pcap_file (required), server_ip, threshold (default 0.02)	A retransmission is a repeated (tuple, sequence, payload-length) data segment. Reports data segments, retransmissions, the rate, whether it exceeds the threshold, and the top offending connections.
analyze_traffic_flow	pcap_file (required), server_ip (required), server_port	Packets and bytes to and from the server, the unique client count, and the top clients by bytes.
analyze_sip_packets	pcap_file (required)	SIP requests by method, responses by status class, and calls (Call-ID, From, To). Payloads on port 5060 plus any payload starting with a SIP method or SIP/2.0 are considered.
analyze_capinfos	pcap_file (required)	Packet count, duration, data bytes, average packet size, first/last packet times, encapsulation, and capture length.

Scope and limits

The parser covers classic libpcap only (big/little endian, micro- and nanosecond timestamps) — no pcapng. Only IPv4 is decoded; IPv6 and other ethertypes are counted in packet totals but not analysed. There is no TCP stream reassembly, so SIP over TCP is matched per segment, and TLS payloads are opaque. Anomaly and state heuristics are deliberately simple approximations, not a substitute for a full tshark-grade analysis.

Development

npm install
npm run build   # tsc -> dist/
npm test        # node test runner over test/*.test.ts, synthetic captures only, no network
npm run dev     # run the CLI from source via tsx

License

MIT — see LICENSE.