Open Source Dec 24, 2025 5 min
AutoAgents: Build Agents in Rust
How to build production-ready agents in Rust with AutoAgents: providers, tools, and scalable runtimes.
Read articleLiquidOS
Building Blocks for the
Agentic Future
AutoAgents is our open-source Rust SDK for building reliable agent systems today. LiquidOS Platform is coming soon as an agentic operating system: local-first for embedded systems, robotics, and local agent execution, with tracing, policy controls, evals, and deployment workflows.
AutoAgents SDK (OSS) Rust-native runtime WASM tool sandboxing
Available now
AutoAgents SDK
Open-source Rust runtime for agent systems.
Coming soon
Platform tracing
Audit-friendly traces, artifacts, and debugging context.
Coming soon
Policy + evals
Budgets, approvals, and regression suites.
Vision
AI for the physical world
Our vision is to enable everyone to interact with AI in the physical world.
Mission
Speed, safety, portability — without tradeoffs
Local-first by default • Embedded + robotics ready •
Measurable outcomes
LiquidOS is for teams that can’t rely on “best effort” AI. Run agents locally on embedded systems and robots, keep execution deterministic, and layer in platform-grade tracing, policy, and evaluation as you scale fleets.
Platform (coming soon)
Modules that make agents shippable
LiquidOS Platform is in active development. Join the waitlist for early access and roadmap previews.
Execution
Local-first runtime
Run where your data lives: embedded devices, robots, and edge nodes without cloud lock-in.
Safety
Policy + approvals
Budgeting, approvals, and scoped tools to align agent behavior with operational rules.
Observability
Tracing by default
End-to-end traces, artifacts, and debugging context for audit-ready deployments.
Evaluation
Evals in the loop
Measure reliability with repeatable evals and feedback loops that improve over time.
Portability
WASM tool sandboxing
Portable, secure tool execution for consistent behavior across machines and environments.
Deployment
Device fleet installs
Fleet deployment patterns with hardened installs and update channels for embedded systems, robots, and edge deployments.
Overview
What is LiquidOS?
A research-driven, production-first effort to make agent systems reliable, observable, and deployable anywhere.
LiquidOS is building a local-first agentic operating system for environments where security, latency, reliability, and observability are non-negotiable — from embedded systems and robots to local execution on edge servers.
AutoAgents is available now as the open-source SDK. The LiquidOS Platform layer is coming soon for teams that need local-first operations, device fleet controls, and reliable agent execution.
Built for local-first environments
- • Secure by default with traceability and auditability
- • Deterministic Rust runtimes with portability baked in
- • Evals, feedback loops, and observability built in
- • Local-first execution across embedded devices, robots, edge nodes, and servers
Workflow
From prototype to deployment
A platform flow that makes “agent in production” a repeatable process.
Step 1
Build with AutoAgents SDK
Compose agents, tools, and memory with structured, testable interfaces.
Step 2
Add platform controls (soon)
Tracing, policy, and evals to ship confidently on embedded systems and robots.
Step 3
Deploy anywhere
Local-first execution across laptops, private servers, edge, and robots.
Why
Why LiquidOS?
A platform stance centered on locality, safety, speed, and measured outcomes.
Local-first
Data stays close
Run agents where your data lives — laptops, private servers, edge devices, and robots. No forced cloud dependency.
Production-grade
Audit-ready from day one
Observability, traceability, and hardened device installs with policy controls and audit logs.
Performance
Fast & portable runtimes
Native Rust runtimes keep overhead low while staying portable across environments.
Research → Production
Measured reliability
Integrated evaluation, tracing, and feedback loops to move confidently from experiments to deployment.
Audience
Who is this for?
Teams building AI systems where reliability and operational clarity are non-negotiable.
Embedded systems teams
Shipping local agents on devices, controllers, or edge nodes.
Robotics teams
Building local-first autonomous systems.
Research teams
Need reproducibility, evaluation, and traceability.
Updates
Latest from the blog
Research notes, system design, and production lessons from building agents.
Join the Waitlist
LiquidOS Platform is launching soon. Join the waitlist for early access, roadmap previews, and deployment onboarding.