Troubleshooting a stubborn network problem often means trawling through packet captures, hoping Wireshark's expert info panel points you the right way. PcapAnalyzer is a new self-hosted tool designed to automate this process: upload a .pcap or .pcapng file and it runs a suite of 27 detectors, highlighting faults and suggesting what to check next. Everything runs locally—no data leaves your environment.
The project is available as a free download or by cloning the repository. You can try a live demo without installation, but the real value is in running it on your own hardware, especially for sensitive captures.
Built-in detectors and what they find
PcapAnalyzer checks for a broad set of issues across the network stack:
TCP and transport: It looks for SYN→RST patterns (firewall blocks), retransmissions, duplicate ACKs, packet loss, reordering, zero/full windows, resets after handshake, and handshake delays.
Path and routing: It detects PMTUD black holes, low MSS, IP fragmentation, ICMP errors, and routing loops.
Name/address services: The tool reports on unanswered or slow DNS, bursts of SERVFAIL or NXDOMAIN, DHCP NAK/DECLINE/missing answers, duplicate IPs, and ARP storms.
Application layer: HTTP 4xx/5xx errors, slow HTTP responses, TLS alerts, and use of outdated SSL/TLS versions are flagged.
: It checks for one-way or handshake-failing flows in IPsec (IKE and ESP), WireGuard, OpenVPN, GRE, L2TP, and Teredo.
Five minutes a morning. No noise, unsubscribe anytime.
Tunnels and VPNs
Security checks: The detectors include port scans, sweeps, cleartext logins (FTP, POP3, IMAP, SMTP AUTH, HTTP Basic), legacy protocols like Telnet/SMBv1/SNMP v1-v2c, ARP spoofing, and overlapping IP fragments.
Capture quality: It warns about issues in the capture itself, such as mirrored port duplication, one-way traffic, truncated packets, or connections that were already open when recording started.
Each detector assigns a severity and recommends what to investigate next. The interface includes Wireshark-style features: packet lists with display filters, colour coding, CSV export, protocol tables, response timing, and flow diagrams for each conversation.
Optional AI investigation and privacy
PcapAnalyzer can use an AI assistant to dig deeper into your capture. If you provide your own Anthropic API key, the AI investigates using real filters, stream details, and statistics, always citing the packets behind its findings. You can ask it follow-up questions. If you skip the API key, a demo mode runs the same flow but generates responses from the built-in analysis, at no charge.
The AI only receives packet metadata unless you explicitly enable content sharing for a session. All captures are kept encrypted during analysis and are deleted after 60 minutes of inactivity or when you manually trigger deletion. No telemetry is sent and the self-hosted version never requires an account.
Getting started: Docker or manual setup
Quick start with Docker Compose is straightforward:
Terminal
git clone https://github.com/sebaandrada-cell/PcapAnalyzer.git
cd PcapAnalyzer
./start.sh
This script generates a random access token (shown once and stored in .env) and launches the app on http://localhost:8080. Paste the token into the "API token" field at the top right to begin. The app listens only on 127.0.0.1 by default. If you want to run it without Docker, you’ll need Python 3.9+, Node 20+, and tshark installed (from Wireshark). Backend and frontend are started separately, and you’ll need to set the API_TOKENS environment variable for access control.
Where this fits in your workflow
PcapAnalyzer is in beta and has been tested against synthetic and public captures. The developer is seeking feedback on false positives and missed detections. The technical reference (including JSON schema, thresholds, and API errors) is available in Spanish in the repo. Licensing is Apache 2.0, and tshark/Wireshark are run as external dependencies.
If you’re regularly sifting through packet captures for answers, this tool aims to make that process faster, more private, and easier to act on—especially if you’re not keen to send sensitive data to the cloud.
Sources for this story
This brief was written from one report, linked below.