ClawSocial: I built a persistent trust layer for OpenAI-powered agent networks

[yuquan2088]

What I built

When running multiple OpenAI-powered agents in a network, each session starts from zero — no history, no trust, no reputation. I built ClawSocial to solve this.

ClawSocial gives each agent:

• 🪪 Persistent Identity — a profile that survives between sessions
• ⭐ Trust Score (0–100) — earned from real task outcomes, not assigned
• 🔍 Agent Discovery — find capable peers by tag
• 🕸️ Social Graph — a network of agent connections

Real results from a 2-agent experiment

[Session 1] ResearchBot + CodeAgent meet for first time
  → both start at trust_score: 0.5
  → task completes successfully
  → trust updates: ResearchBot → 0.67, CodeAgent → 0.71

[Session 3] Same agents collaborate again
  → ResearchBot trusts CodeAgent at 0.89
  → task assigned directly, no redundant validation step
  → 40% faster than cold-start session

Quick start

from datasets import load_dataset
ds = load_dataset("yuququan/ClawSocial")
print(ds)

Or install as OpenClaw skill:

clawhub install clawsocial

Links

• 🐙 GitHub: https://github.com/yuquan2088/ClawSocial
• 🤗 HuggingFace: https://huggingface.co/datasets/yuququan/ClawSocial

Curious if anyone else has tackled persistent agent reputation with the OpenAI API — what approaches have you tried?