Glossary
A comprehensive reference of Meta-OS terminology, concepts, and technical terms.
Agent
A lightweight software component (1-10MB) that extends Meta-OS capabilities. Unlike traditional apps, agents have no UI and provide capabilities that the OS orchestrates.
AI Orchestrator
The core intelligence layer of Meta-OS that understands user intent, coordinates agents, manages context, and enforces security policies.
Capability Declaration
A structured manifest that specifies exactly what an agent can do, in plain language. Includes permissions, API domains, frequency limits, and purpose explanations.
Intent-Based Architecture
Design paradigm where users express what they want (intent) rather than navigating UIs. The OS determines how to fulfill the intent using available agents.
On-Device AI
AI models that run locally on your phone without internet. Meta-OS uses quantized LLMs for voice understanding, intent parsing, and basic reasoning — keeping data private.
Agent Store
Marketplace for discovering and installing agents. Supports voice discovery ('Find me a photo editor agent') and visual browsing with ratings and reviews.
Sandbox
Isolated execution environment where agents run. Prevents direct system access, enforces resource limits, and ensures agents can't interfere with each other.
Meta Assistant
The built-in conversational AI interface of Meta-OS. Your primary way to interact with the system via voice or text.
Personal Memory
AI-powered note-taking and memory vault. Stores ideas, meeting notes, and observations with automatic categorization and natural language search.
Orchestration
Process where the AI Orchestrator invokes multiple agents in sequence to complete complex tasks. Example: 'Book dinner after the movie' triggers Calendar → Maps → Restaurant agents.
Capability
Specific permission an agent requests (e.g., location.read, calendar.write, network.fetch). More granular than traditional app permissions.
Agent Lifecycle
The stages an agent goes through: Discovery → Installation → Registration → Invocation → Execution → Update → Removal.
Code Signing
Cryptographic signature applied by agent developers. Meta-OS verifies signatures to ensure agents haven't been tampered with.
Audit Log
Complete record of all agent activity: what data was accessed, when, and for what purpose. Accessible in Settings → Privacy → Activity Log.
Voice Discovery
Finding agents through natural language. Say 'I need a fitness tracker' and Meta-OS recommends relevant agents from the store.
Context Management
How the AI Orchestrator maintains conversational state. Remembers references like 'the movie,' 'Sarah,' 'tomorrow' across multiple interactions.
Verified Developer
Developer or company that's verified their identity with Meta-OS. Verified agents show a badge for enhanced trust (e.g., Spotify, Uber).
Resource Limits
Constraints on agent execution: maximum CPU time, memory usage, network bandwidth. Prevents malicious agents from draining battery or data.
Least Privilege
Security principle where agents receive only the minimum permissions needed. Example: Access only one contact, not all contacts.
Agent Registry
System database where installed agents register their capabilities. The Orchestrator queries this registry to determine which agent to invoke for a given intent.
Expo Router
React Native navigation framework powering Meta-OS's seamless app transitions and deep linking.
Intent Resolution
Process of parsing natural language input, extracting entities (people, times, locations), and mapping to agent capabilities.
Auto-Update
Background update mechanism for agents, similar to Chrome extensions. Updates install silently unless new capabilities require user approval.
Secure Enclave
Hardware-isolated area of the processor used for storing encryption keys and biometric data. Ensures sensitive data never enters main memory.
Data Portability
Ability to export your data in standard formats (JSON, CSV). Meta-OS provides built-in export for calendar, contacts, Personal Memory, and agent data.
TLS 1.3
Transport Layer Security protocol used for encrypting all network communication between Meta-OS and cloud services.
Quantized LLM
Large Language Model optimized for mobile devices by reducing precision (e.g., 8-bit instead of 32-bit). Enables fast on-device AI with lower memory footprint.
Permission Revocation
Instantly removing agent access to capabilities. Changes take effect immediately without restart. Say 'Remove Weather Agent permissions' or use Settings.
API Domain Whitelist
List of approved domains an agent can access via network.fetch capability. Example: Weather Agent can only call api.weather.com, not arbitrary URLs.
Agent Manifest
JSON file included with every agent that declares metadata (name, version, developer), capabilities, and update URL.
Can't find what you're looking for? Join our community or check out the API reference.