What is a Governed Autonomous System (GAS)?
A Governed Autonomous System (GAS) is an AI agent architecture where every action is constrained by a constitutional governance document (the Soul), classified by risk tier (T0-T3), verified after execution, and recorded in an immutable audit trail. Unlike standard agent frameworks, governance is architectural, not advisory. The agent cannot act outside the boundaries defined by its Soul, regardless of what the underlying model reasons.
How does Lancelot differ from LangGraph?
LangGraph defines how an agent works through graph-based state machines. Lancelot defines what an agent cannot do through constitutional constraints. LangGraph has no Soul, no Risk Tier, no Trust Ledger, no immutable receipt system, and no local PII scrubbing. Lancelot is governance-first: capability is progressively enabled on top of a constitutional foundation. For applications that touch production infrastructure, customer PII, or require compliance export (SOC 2, ISO 27001, GDPR), Lancelot provides architectural guarantees that LangGraph cannot.
How does Lancelot differ from CrewAI?
CrewAI assigns roles to orchestrate agent collaboration and defines how agents work together. Lancelot defines constitutional constraints on what any agent can do, regardless of its role. CrewAI has no constitutional governance, no risk-tiered pipeline, no Trust Ledger, and no immutable audit trail. Lancelot's Hive Agent Mesh enforces monotonic restriction: sub-agents can only be more restricted than their parent, never less. This makes Lancelot suitable for enterprise deployments requiring compliance, auditability, and behavioral guarantees.
Does Lancelot support SOC 2 and ISO 27001 compliance?
Yes. Lancelot includes a Compliance Export subsystem that generates one-click SOC 2 Type II, ISO 27001, and GDPR Article 30 reports directly from the immutable receipt DAG. Every action, approval, risk classification, and verification result is permanently recorded in structured receipts with chain integrity verification. The receipt trail is the compliance evidence.
What is the Trust Ledger?
The Trust Ledger is Lancelot's progressive trust graduation system. Rather than treating agents as binary (trusted or untrusted), the Trust Ledger tracks every action outcome. After 50 consecutive successes for a given capability, the system proposes a tier reduction, requiring less oversight for proven behavior. A single failure triggers instant revocation. Trust is earned slowly and lost immediately.
Does Lancelot protect against prompt injection?
Yes. Lancelot's Soul constraints are architecturally immune to prompt injection. The governance layer uses 16 banned pattern detectors, 10 regex detectors, homoglyph normalization, and zero-width character stripping at the input layer. Because the Soul is a versioned constitutional document that cannot be modified without owner approval, a prompt injection attack cannot alter behavioral boundaries at runtime.
What is the Universal Application Bridge (UAB)?
The Universal Application Bridge (UAB) gives AI agents framework-level control over desktop and server applications without APIs, plugins, or screen reading. It hooks directly into application frameworks (Chrome DevTools Protocol for Electron apps, COM Automation for Microsoft Office, UI Automation for Windows apps) to enumerate UI elements, read state, and execute actions on real controls. UAB supports 9+ UI frameworks, 61 action types, 17 native MCP tools, and includes a Spatial Map Engine for 2D spatial understanding of application interfaces.
How do I install Lancelot?
Run: npx create-lancelot. The installer performs a 13-point pre-flight check covering Node.js, Docker, disk space, RAM, GPU, network, ports, and permissions before installing anything. It guides you through selecting communication connectors (Gmail, Slack, Telegram) and AI providers (Anthropic, OpenAI, Google, xAI), then opens the War Room dashboard at localhost:8501. No manual configuration required.
