EnterpriseOverviewIntelligenceWho It's ForSecurityIT Governance
Pricing
ResourcesBlogFAQ
Get in touch
How it works
Four layers. Enterprise controls at every boundary.
The Studio is a four-layer desktop application: React UI, Electron process, Rust sidecar, and cloud server. Security-sensitive operations never run in the UI layer. Policy enforcement starts before the workspace loads.
Startup
Enterprise controls activate before the workspace loads.
Authentication with SSO support (SAML, OIDC). Organisation policy fetched and cached for offline enforcement. Legal acceptance gates for terms and compliance. Managed device detection for enterprise deployments. The workspace only renders after all gates pass. Updated terms can appear mid-session without interrupting work.
Agents
Every agent is configured with a vendor, model, profile, role, and tool set.
Agents live in slots within a project. Each slot specifies an AI vendor, a model, a personality profile (instructions, language, tone), a role (remit, allowed tools, file path restrictions), and reusable trait blocks (personas, skills, constraints, workflows). Kits deploy a pre-configured team of agents in one action.
10+ AI providers supported: Anthropic, OpenAI, Google, Deepseek, Groq, Mistral, Bedrock, Azure, Databricks, and local inference via Ollama. Organisation policy restricts which vendors and models are available.
Multiplex
A conductor generates the plan. Specialists execute. Handoffs carry context forward.
In a team session, a conductor agent creates an execution plan with steps assigned to specific agents. Each step runs with approval gates. When one agent completes its work, the conductor composes a context package and hands off to the next agent. The human can intervene, edit the plan, or skip steps at any point.
The plan tracks progress visually. Token costs aggregate across all team members. Session history captures the full multi-agent collaboration for replay and audit.
Architecture
Sensitive operations run in a separate secure process, isolated from the interface.
A dedicated local process handles everything security-sensitive: data masking, encryption, policy enforcement, tool routing, analytics, and voice processing. The user interface never touches raw sensitive data. If anything goes wrong, the secure process restarts automatically without losing work in progress.
This separation means that even if the application has a vulnerability, protected data stays protected. The secure process monitors its own health continuously and recovers from failures without user intervention.
Traffic control
Every AI request intercepted. Every tool call routed through a single gateway.
The transform proxy intercepts all LLM API calls. Data Shield masking is applied, policy compliance verified, and the request is audit-logged before forwarding to the vendor. Responses are unmasked on return. Per-vendor usage statistics track cost and volume.
The MCP gateway multiplexes tool calls between agents and providers: plugin tools, MCP servers, and terminal operations. Each agent slot has its own tool preferences. Results are cached where appropriate. The gateway runs on the local Hive sidecar, so all tool routing happens on the local machine.
See how Swarmix works
A walkthrough of unified visibility, governance, and intelligence across every AI vendor and team.
- Deploy in weeks
- Works with any existing AI stack
- SOC 2 compliant, GDPR ready