Index / Loong · updated Oct 2, 2026
Loong
loongclaw-ai/loongclaw · healthy · rank 30 of 43 by stars
A Rust-native foundation for vertical AI agents with an architecture-first, policy-authorized security model. Currently an early skeleton rewrite — promising design, but not yet a usable agent product.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 72 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 69 files examined
6 providers · gateway support
Pinned models
deepseek-chat released 2024-12-26 — 22 months old (from the public model catalogue)
Pin last edited 6 months ago
Evidence
Decision
Why choose Loong over OpenClaw?
Why choose this
- Rust core offers lower runtime overhead and memory footprint
- Explicit Action authorization model before execution
- Actor runtime (loac) with clean lifecycle management
Tradeoffs
- Not a usable product yet — skeleton stage only
- No tool integrations, channels, or UI
- No releases, docs for end users, or community traction
Best fit
- Rust developers building custom vertical AI agents
- Teams wanting a policy-first security architecture
- Contributors interested in actor-model agent runtimes
Avoid if
- You need a working agent product today
- You want a mature plugin/tool ecosystem
- You prefer batteries-included frameworks like OpenClaw
README and commits clearly show an active but early-stage architectural rewrite, so maturity assessments are high-confidence. Community sentiment is uncertain due to zero Reddit mentions and only indirect web ecosystem coverage.
AI layer reviewed Sep 21, 2026 · how this is written
Star activity
640 stars today
Overview
Loong is a Rust-based foundation for building vertical AI agents, positioning itself as a security- and performance-first alternative in the OpenClaw ecosystem. Its core architectural idea is architecture-governed safety: outbound effects are modeled as Actions that must be authorized by policy before execution, with sandboxes and model review treated as defense-in-depth rather than the primary boundary.
Under the hood, Loong builds on its own actor runtime, loac, which provides a lifetime model making components easy to run, extend, stop, and restart. Recent commits show heavy investment in converging the actor message API (reply/stream shapes, sync handlers, dispatch strategies) and building agents from serializable AgentConfig, indicating the framework plumbing is maturing even though product features are not.
Compared to OpenClaw, Loong is explicitly not ready to install or use — the maintainers describe it as an independent-history rewrite containing an architecture and contract skeleton. Its long-term bet is that a small, well-governed Rust core with dual extension paths (compiled Rust integrations plus runtime scripts) will out-evolve larger but less disciplined codebases. For now, it is best suited to contributors and Rust developers who want to shape an agent framework from the ground up rather than end users seeking a working assistant.