FREE VS CODE EXTENSION
v0.1.5 · MIT License

NetTrace Agentix

Drop in a .pcap file, type @nettrace what's wrong? in Copilot Chat, and get expert-level network diagnosis — no Wireshark expertise required.

Requires GitHub Copilot Requires Wireshark / tshark VS Code 1.95+
Copilot Chat
✓ capture.pcapng parsed — 12,847 packets, 4 anomalous streams
YOU @nettrace why are connections to the API dropping?
AI
→ Calling GetTCPStreamDetail(stream: 7)... → Calling GetExpertInfo(filter: "tcp.analysis")... Stream #7 shows TCP RST after every TLS Client Hello. The server is rejecting on port 8443 — likely a firewall rule blocking TLS 1.2. Streams on port 443 complete normally. Recommendation: Check firewall rules for port 8443 or switch the client to port 443.

Wireshark-powered. Copilot-driven.

NetTrace Agentix brings the packet-level depth of Wireshark together with the reasoning power of GitHub Copilot — all inside VS Code. Diagnose network issues through natural conversation instead of manual filter hunting.

Parses Real Captures

Reads .pcap, .pcapng, and .cap files directly. Uses tshark under the hood — the same engine as Wireshark — to extract packets, conversations, and expert diagnostics.

AI That Drills Deeper

The AI doesn't just read a summary — it autonomously calls tools to follow TCP streams, apply display filters, and inspect expert info, just like a senior engineer would.

Natural Language Interface

No filter syntax. No protocol deep-dives required. Just ask @nettrace in Copilot Chat the same way you'd ask a colleague, and get plain-English diagnosis.

Everything You Need to Diagnose Network Issues

Interactive Capture Viewer

A rich webview panel showing the packet list, conversation statistics, and TCP stream summaries — all sorted by anomaly score so the most suspicious connections surface first.

Packet List Conversations Display Filters
Agentic Tool Calling

The AI autonomously calls tshark tools during analysis — drilling into specific streams, following TCP payloads, comparing captures, and applying filters without any manual prompting.

Get Stream Detail Follow TCP Stream Expert Info
Client vs. Server Comparison

Load simultaneous captures from both ends of a connection. The AI correlates packets between captures to find what's missing, delayed, or modified in transit — invaluable for firewall and proxy debugging.

Packet Correlation Drop Detection RTT Delta
Specialized Analysis Agents

Switch to a purpose-built agent for deeper expertise. Each agent has a tailored persona, curated tool set, and pre-applied display filters.

TLS/SSL Specialist DNS Troubleshooter VoIP/SIP Analyzer
Knowledge Templates

Inject domain expertise into every analysis. Add markdown files covering known firewall quirks, Windows TCP stack behaviors, security heuristics, and false positives — hot-reloaded with no restart needed.

Known Issues Security Heuristics Custom Wisdom
Scenario Context

Tell the AI the story before analysis starts. Define the symptom, client/server IPs, topology, and notes. Every prompt is enriched with this context for more targeted, relevant diagnosis.

Symptom Description IP Topology JSON Config

Powerful Commands, Plain Language

Use the @nettrace chat participant with slash commands for focused analysis, or just ask in plain English — the AI figures out what tool to call.

Full Command Reference
@@nettrace what's wrong? General AI diagnosis of the active capture
@@nettrace /summarize Traffic summary with stats and anomaly highlights
@@nettrace /diagnose Root cause analysis using scenario context
@@nettrace /stream 5 Deep dive into a specific TCP stream by index
@@nettrace /compare Compare client-side and server-side captures
@@nettrace /agent tls-specialist Switch to a specialized analysis agent

Up and Running in Minutes

Four steps from zero to AI-powered network diagnosis.

1
Install Wireshark

Download from wireshark.org. tshark is included. Verify with tshark --version.

2
Install the Extension

Search NetTrace Agentix in VS Code Extensions, or install directly from the Marketplace.

3
Drop a Capture File

Drag a .pcap, .pcapng, or .cap file into your workspace. The extension auto-parses it and shows it in the NetTrace sidebar.

4
Ask Copilot

Open Copilot Chat and type @nettrace what's wrong with this capture? — then follow the AI's analysis and drill-down suggestions.

Built for Real-World Scenarios

From quick triage to deep root-cause analysis — any traffic type, any environment.

TLS / HTTPS Issues

Diagnose certificate errors, cipher suite mismatches, failed handshakes, and unexpected RST packets on port 443 or 8443. Switch to the TLS Specialist agent for deep-dive analysis.

DNS Problems

Surface NXDOMAIN floods, slow lookup times, delegation failures, and SERVFAIL responses. The DNS Troubleshooter agent has built-in heuristics for common resolver issues.

Packet Loss & Retransmissions

Identify TCP retransmission storms, duplicate ACKs, zero-window probes, and out-of-order sequences. Anomaly scoring surfaces the worst streams automatically.

VoIP / SIP Quality

Analyze SIP call flows, RTP quality metrics, registration failures, and codec negotiation issues with the VoIP/SIP Analyzer agent.

Security Analysis

Detect suspicious scan patterns, unexpected beaconing, credential-in-clear traffic, and port sweep anomalies using the built-in security knowledge templates.

Cloud & Proxy Debugging

Compare client and server captures through a corporate proxy or cloud load balancer to find exactly where packets are being lost, modified, or delayed.

No Extra API Keys. No External Services.

NetTrace Agentix uses vscode.lm.selectChatModels() — the same models your GitHub Copilot subscription already provides. There is no separate backend, no cloud service, and no telemetry beyond what VS Code itself collects.

All capture parsing runs locally via tshark on your own machine. Your network data never leaves your environment.

Fully local — captures stay on your machine
No API key required beyond Copilot
Open source under MIT license
// Three-Layer Architecture
Parsing Layer
→ tshark runs locally on your .pcap files
→ Parallel execution for speed
Context Assembly Engine
→ ~900K token budget management
→ Anomaly-score ranked streams first
→ Domain knowledge injection
Presentation Layer
→ @nettrace Copilot Chat participant
→ Capture Viewer webview
→ LM Tools (AI calls these autonomously)

Start Diagnosing Networks with AI

Free, open source, and built by the team at Cognitive Agentics as a demonstration of how AI can be embedded deeply into developer workflows. If you'd like a similar AI-powered experience built into your own infrastructure or tooling, we'd love to talk.