Features
5-10x faster inference via native vLLM integration with auto multi-GPU and PagedAttention
Grounded citations: response.sources/.citations filled from retrieved URLs
Reasoning model support: gpt-5, o-series complete with cost/token/latency reporting
One-call domain agent creation: LegalDomain().to_agent()
Custom personas enforced on every execution path
Fail-closed agent.run(): never returns success with empty output
Self-updating model catalog with drift warnings
Sandboxed built-in tools with SSRF guard and path-confined file tools
Policy-based ModelRouter: FirstAvailable, CostBased, LatencyBased with transparent failover
Automatic task decomposition and sub-agent routing (AgentMode.AUTO)
Multi-agent orchestration with team patterns, shared state, and message bus
14 inference backends: 5 local engines + 9 cloud providers (OpenAI, Anthropic, Gemini, Cerebras, Groq, Together,
Unified ProviderRegistry with list_providers()/list_models() and API readiness checks
66+ built-in tools covering computation, code execution, web search, documents, media, and more
9 agent presets: math, research, coding, general, rag, media, notify, multimodal, minimal
Laguna XS.2 model (33B params, 3B active, on-device)
Laguna M.1 model (225B params, 23B active, via API)
256K context length for multi-step reasoning
Multi-agent orchestration with planning and tool use
Sandboxed agent execution environments
Developer surfaces: agents, TUI, IDE extensions, binaries
Custom foundation models deployed on-prem or in VPC
Data connectors to repos, databases, data warehouses
Role-based access control for humans and agents
Executive-grade governance and auditability
On-prem, VPC, and workstation deployment (defense only)
Forward Deployed Research Engineers embed with teams
Real-time observability with end-to-end traces
Air-gapped network support
Open-weight model weights available
