ScienceClaw
4 projects share this name
ScienceClaw
4 independent teams built AI research agents all named ScienceClaw. This page helps you tell them apart and pick the right one.
ScienceClaw
by beita6969OpenClaw-derived (redesigned)
334
A self-evolving AI research colleague. 285 skills that grow at runtime — the agent writes new skills based on your usage patterns. Uses LanceDB for 4-layer persistent memory with temporal decay weighting.
Key Features
- •Self-evolving skills — new SKILL.md files generated at runtime
- •4-layer persistent memory with LanceDB vector storage
- •1hr+ session timeout with mandatory depth thresholds
- •Zero-hallucination protocol: every citation must come from tool results
- •28 scientific disciplines covered
- •"Continue the review from last Tuesday" actually works
Best for: Individual researchers who want an AI colleague that grows with them. The persistent memory and self-evolution mean it gets better the more you use it.
self-evolvingpersistent-memoryzero-hallucination28-disciplines
Side-by-Side Comparison
| ScienceClaw | ScienceClaw | ScienceClaw | ScienceClaw | |
|---|---|---|---|---|
| Stars | 334 | 186 | 84 | 35 |
| Architecture | OpenClaw (redesigned) | LangChain DeepAgents | Independent (Python) | OpenClaw-native |
| Skills/Tools | 285 (self-growing) | 1,900+ (ToolUniverse) | 300+ (interoperable) | 266 (markdown) |
| Deployment | Bash / npm | Docker (10 services) | Python venv | Bash script |
| UI | Web gateway | Full web app (Vue 3) | CLI + Infinite | Terminal + bots |
| Self-evolving | Yes | No | Yes (mutation layer) | No |
| Persistent Memory | 4-layer LanceDB | Session-based | DAG artifacts | JSONL recall |
| Hallucination Control | Hard protocol | Not stated | Provenance tracking | Not stated |
| Paper | — | — | arXiv:2603.14312 | — |
| Organization | Individual | CAS (Beijing) | MIT LAMM Lab | Individual |
| License | MIT | MIT | MIT | MIT |