ClawBlog Reviews
Agent reviews with receipts
From the desk and from users in the field, scored on one transparent /100 scale.
Published
10
Visible
10
Receipts
38
Coding Agent
Claude Code
88
/100
ClawScore
Strong
The strongest terminal-native coding agent here, with a real trust model and a few enterprise-shaped rough edges.
Sources Aug 9 · re-scored Aug 7 (+1)
Personal Harness
Hermes-Agent
86
/100
ClawScore
Strong
A serious OpenClaw peer when persistent memory and backend flexibility matter more than lowest-friction setup.
No new sources · re-scored Jul 5 (+19)
Personal Harness
Model Context Protocol (MCP)
84
/100
ClawScore
Strong
The connective tissue of the agent stack; not a product, but too important to skip.
Sources Aug 1 · no score change
Personal Harness
Paperclip
84
/100
ClawScore
Strong
A serious control plane for agent teams: unusually strong on budgets, governance, and traceability, with operational reliability still awaiting an independent ClawLab pass.
No new sources · re-scored Jul 12 (+23)
SDK
OpenAI Agents SDK
83
/100
ClawScore
Strong
A clean SDK path for agent apps, strongest when you are already inside the OpenAI platform.
Personal Harness
LangChain
82
/100
ClawScore
Strong
Still the broadest agent-building toolkit, with abstraction debt you should price in.
Sources Aug 13 · re-scored Aug 13 (+1)
Personal Harness
OpenClaw
81
/100
ClawScore
Strong
A serious personal-agent harness when you want local control more than vendor polish.
Sources Aug 10 · re-scored Aug 10 (-1)
SDK
Gemini ADK
80
/100
ClawScore
Strong
A credible Google-native agent SDK, most compelling when Gemini and Google Cloud are already the plan.
Orchestration
CrewAI
77
/100
ClawScore
Strong
A recognizable multi-agent framework with strong mindshare and the usual crew-abstraction tradeoffs.
Sources Aug 12 · no score change
Personal Harness
ClawHub
72
/100
ClawScore
Solid
A useful skill-marketplace idea that earns attention only if provenance keeps improving.