Strike48 components run in three deployment models. Same connectors, same tools — what changes is where they run and how they talk to each other. The diagrams below are interactive: drag to pan, use the toolbar to zoom, click any node to open its docs, or go fullscreen for a closer look.
At a Glance
| Model | Who it's for | Transport | Topology |
|---|---|---|---|
| Desktop | Individuals, local work | Unix domain socket (IPC) | One workstation |
| Platform | Teams, remote execution | gRPC + mTLS | Hub + hosted connectors |
| Distributed | Campaigns across networks | gRPC + mTLS | Many agents, central control |
Desktop Model
You run StrikeHub on your machine; it launches and supervises connectors locally. Everything happens on one host, and connectors talk to the shell over Unix domain sockets — nothing is exposed to the network.
- Single window — every connector gets a panel, all sharing one app and one OIDC login.
- No network exposure — IPC over Unix domain sockets means connectors aren't reachable off-host.
- Local credentials — KubeStudio uses your local kubeconfig; results stay on your machine.
Platform Model
Connect StrikeHub — or your connectors directly — to Prospector Studio, the hosted backend. Now AI agents and workflows can route jobs to your connectors over gRPC, with mutual TLS and centralized audit logging.
- Remote execution — trigger tools via the GraphQL API or workflows.
- Team collaboration — shared engagements, knowledge bases, and a centralized audit trail.
- mTLS everywhere — connector-to-platform traffic is mutually authenticated; users authenticate via OIDC/OAuth2 with RBAC on every job.
Distributed Model
The platform model, scaled out: many connector agents deployed across different networks, all registered with one Prospector Studio. An agent on each target network executes locally; results aggregate centrally. This is how distributed scanning and multi-network campaigns run.
- Deploy agents anywhere — each connector registers with the platform and receives jobs over gRPC.
- Aggregate results — findings from every agent roll up into one shared intelligence database.
- Orchestrate campaigns — drive distributed scanning from a single workflow.
Pick remote execution → · Deploying connectors →
Mixing Models
Connectors are independent processes, so you don't have to pick one model for everything. Run KubeStudio locally through StrikeHub while a fleet of Pick agents runs in the platform — they don't need to know about each other. See the Hub & Connector Model for the full picture of communication patterns and security boundaries.
Choosing a Model
Next Steps
- Individuals: Download StrikeHub and launch connectors locally.
- Teams: Contact the Strike48 team for Prospector Studio access, then configure your connectors.
- Developers: Build a connector with the SDK and deploy it.