Astrid

Astrid is a portable user-space runtime for WebAssembly capsules that isolates agent systems, enforces capability-based access, and manages durable state.

Astrid is a portable runtime that manages software built from WebAssembly capsules. It enforces security by separating agent intelligence from system authority, using a capability-based model where files, network access, and IPC are checked at runtime boundaries. Each capsule runs in a WebAssembly sandbox with no ambient authority, importing only declared capabilities via a typed ABI. The system is designed for agent frameworks and services, ensuring that components communicate through an event bus with versioned interfaces. It maintains durable state for principals, providing isolated storage and secrets that remain secure even when capsules are updated or removed during runtime. The runtime currently runs on macOS and Linux.

Astrid at a glance

Best for: Software engineer
Runs on: macOS, Linux, CLI, self-hosted
Pricing: Free
Released: February 2026
Category: AI agent tool
Website: github.com/astrid-runtime/astrid

Key features

Cryptographic capability model for all resources
WebAssembly sandbox without ambient authority
Typed interfaces for inter-capsule communication
Content-addressed storage for portable state
Signed, hash-linked audit chains
Live installation and upgrading of capsules
Per-principal isolation for secrets and data

Use it for

Building agent systems with isolated permissions
Creating secure service distributions
Hosting independent software components
Managing agent state and secrets

Alternatives to Astrid

OpenViking — Unifies agent memory, knowledge RAG, and skills into a single directory structure.
AgentDock — It coordinates multi-device workflows and provides secure environment access.
Agent Sandbox — Provides a secure and isolated execution layer for autonomous AI agents.
Eve — Builds durable AI agents with file system based configuration and isolated sandboxes.
Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
BrowserOS — Runs entirely locally and lets AI agents use real user logins to complete tasks.
Sandcastle — Orchestrates AI coding agents in isolated Docker, Podman, or Vercel sandboxes.
AgentFS — Provides isolated filesystem access and instant snapshots via SQLite.
synapseForge — An open source framework for autonomous agents with local execution and desktop distribution.
Agent Delegated Authority — It defines cryptographically verifiable authorization receipts and deterministic boundaries.
Atomic Agent — It runs locally on your machine, ensuring data privacy and no token costs.
AgentTrust — Enforces zero-trust policy rules and maintains an immutable cryptographic audit ledger.

All Astrid alternatives