Here's the short version: we build tools for security and infrastructure teams, and we connect them with AI that actually does useful things. Not "AI-powered" as a marketing checkbox — AI agents that talk to your data, run your tools, and give you answers.
The slightly longer version involves three layers that talk to each other.
The Three Layers
Prospector Studio — the brains
This is the backend. Workflow engine, agent runtime, GraphQL API, and a web-based studio where you wire things together. Think of it as the place where AI agents live, knowledge bases get built, and workflows run. It's also where connectors register themselves so agents know what tools are available.
StrikeHub — the desktop shell
You could run everything through a browser, but we figured a native desktop app would be nicer. StrikeHub is a single window that hosts all your connectors side by side — each one gets its own panel, but they share auth, notifications, and connectivity to Prospector Studio. One app, all your tools.
Connectors — the tools that do the work
Each connector is a standalone tool focused on one job:
- KubeStudio — Kubernetes cluster management. Browse namespaces, inspect workloads, tail logs, manage deployments. All from your desktop.
- Pick — Penetration testing toolkit. Network scanning, device discovery, traffic capture, remote execution. Lightweight and headless.
- StrikeKit — Red team engagement management. Plan, execute, and report on pen test engagements with AI assistance.
Connectors run as separate processes — if one crashes, nothing else goes down. They serve their UI over HTTP and talk to StrikeHub through Unix domain sockets, so nothing is exposed to the network.
How They Talk to Each Other
The pattern is simple: connectors register their capabilities with Prospector Studio over gRPC. When an AI agent needs to scan a network or check a Kubernetes deployment, it knows which connector to call and how to call it. StrikeHub handles the desktop side — launching connectors, managing their lifecycle, bridging WebSocket connections.
The result: you ask a question in natural language, an agent figures out which tools to use, calls them, and assembles the answer. Or you skip the AI entirely and use the tools directly. Your call.
Want the deep dive?
The Hub & Connector Model page covers the full architecture — communication patterns, deployment models, security boundaries, and how to access Prospector Studio from different environments.
We build in Rust. Here's what's under the hood:
| Component | Technology |
|---|---|
| Language | Rust (2024 edition) |
| UI Framework | Dioxus 0.6 |
| WebView | Wry |
| HTTP Server | Axum 0.7 |
| Async Runtime | Tokio |
| IPC Transport | Unix domain sockets |
| Platform Integration | gRPC (Prospector Studio) |
| Authentication | OIDC / OAuth (Keycloak) |
| Configuration | TOML |