AgentOS Documentation

Modular orchestration runtime for AI agent systems.

npm install @framers/agentos

Quick Navigation

Start Here

• High-Level API — generateText(), streamText(), generateImage(), and agent()
• Documentation Index — Installation, architecture, and core guides
• TypeDoc API — Generated API reference for the full runtime

Getting Started

• Documentation Index — Installation and quick start
• High-Level API — generateText(), streamText(), generateImage(), and agent()
• Ecosystem — Related packages and resources
• Releasing — Version history and release process

Architecture & Core

• System Architecture — System design and core internals
• Platform Support — Supported environments
• Tool Calling & Lazy Loading — Extension loading, schema-on-demand, and descriptor IDs
• Observability (OpenTelemetry) — Traces, metrics, and OTEL-compatible logging (opt-in)
• Logging (Pino + OpenTelemetry) — Structured logs, trace correlation, and OTEL LogRecord export (opt-in)

Planning & Orchestration

• Unified Orchestration Layer — One runtime, three authoring APIs
• AgentGraph — Explicit graph builder with node/edge control
• workflow() DSL — Deterministic DAG pipelines
• mission() API — Goal-first orchestration
• Checkpointing — Resume, replay, and forking
• Planning Engine — Multi-step task planning
• Human-in-the-Loop — Approval workflows
• Agent Communication — Inter-agent messaging
• Multi-Agent Agency API — agency(), strategies (sequential, parallel, graph, hierarchical), dependsOn

Safety & Security

• Guardrails — Content filtering, PII redaction, and folder-level filesystem permissions
• Creating Custom Guardrails — Building and composing guardrail pipelines
• Safety Primitives — Circuit breakers, cost guards, stuck detection, and tool execution guards
• Provenance & Immutability — Signed event ledger, soft-delete tombstones, revision history, and autonomy guard
• Immutable Agents — Toolset pinning, secret rotation, and soft-forget memory patterns

Memory & Storage

• Cognitive Memory — Personality-modulated memory, retrieval, and consolidation
• Working Memory — Markdown notes and Baddeley slot-model working memory
• RAG Memory — Vector storage and retrieval
• Multimodal RAG — Image and audio embeddings
• SQL Storage — Database adapters
• Client-Side Storage — Browser and local persistence

Capabilities & AI

• Capability Discovery — Tiered semantic discovery (~90% token reduction)
• Emergent Capabilities — Runtime tool creation, sandboxed execution, and LLM-as-judge verification
• Deep Research — Multi-source research pipeline with query classification
• Structured Output — JSON schema validation
• Evaluation Framework — Testing and benchmarks
• Cost Optimization — Token usage and caching

Advanced

• Recursive Self-Building — Self-modifying agent patterns
• Agency Collaboration — Multi-agent coordination

Voice & IVR

• Voice Pipeline — End-to-end voice conversation architecture, VAD, barge-in, and turn detection
• Speech Providers — Full catalog of STT, TTS, VAD, and wake-word providers
• Telephony Providers — Twilio, Telnyx, Plivo webhook setup and call management

Skills

• Skills Overview — SKILL.md format, loading, and semantic discovery integration
• Skills Format — Authoring SKILL.md files
• Skills Extension — Installing and managing skill packs
• Skills Registry — Browsing and installing curated skills

Extensions

• Extensions Overview — Available extensions catalog
• How Extensions Work — Loading and lifecycle
• Extension Architecture — Building custom extensions
• Auto-Loading — Automatic extension discovery
• Extension Standards (RFC) — Interface contracts and versioning
• Safety: PII Redaction, Code Safety, Grounding Guard, ML Classifiers, Topicality
• Research: Web Search, Web Browser, News Search
• Media: Voice Synthesis, Image Search, Giphy
• Integrations: Auth, Telegram, CLI Executor

API Reference

• TypeDoc API — Auto-generated API documentation