Runtime

Runtime Subsystems

Detailed breakdown of the core components powering FemtoClaw’s deterministic execution authority. Each subsystem plays a crucial role in enforcing agentic sovereignty and safe AI execution【93068253089653†L245-L269】.

Agent Core

The orchestrator of deterministic execution. The Agent Core manages task scheduling, ensures progress and upholds all runtime invariants. It is the authoritative core of FemtoClaw’s runtime.

Brain Interface

Bridges generative inference systems (LLMs or other models) to the deterministic runtime. It translates high‑level intents into formal tool calls while preserving the strict separation between inference and authority【93068253089653†L233-L241】.

Capability Gate

Implements a deny‑by‑default policy for tool execution. Every system call is capability gated to enforce safety and restrict agents to explicitly authorised operations【93068253089653†L245-L269】.

Protocol Validator

Verifies JSON tool call structures according to the normative specification. Rejects any ambiguous or natural language instructions and guarantees determinism in the execution layer【254469092789682†L3-L29】.

Memory Subsystem

Provides write‑ahead logging (WAL) durability, crash recovery and deterministic state replay for long‑running agents. Every state mutation is persisted before execution【93068253089653†L245-L269】.

Observability Layer

Emits structured logs and metrics for each decision and system call, enabling full auditability and traceability of agent reasoning and hardware interactions【93068253089653†L245-L269】.