Hollow
AI Agents & AssistantsA local, self‑modifying LLM agent platform where three autonomous agents collaborate, learn, and request changes in an unsupervised environment.
Features
- Runs three locally hosted LLM agents that autonomously set goals and interact with each other
- Agents can write their own Python tools via `synthesize_capability` and submit change requests for operator review
- Built‑in stress‑monitor (`suffering`) that locks capabilities when load thresholds are exceeded
- Multi‑layer validation (file substance, modify‑intent, semantic check, codebase fact‑check, peer feedback) before actions take effect
- Persistent workspace artifacts survive restarts; broken items are cleaned up automatically
Recent releases
View all 52 releases →
Review required
v5.5.5
Breaking risk
Auth
Breaking upgrade
RCE / SSRF
Atomic goal execution + capability validation
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Install & Platforms
Install via
pip
docker
Platforms
windows
macos
linux