W&B MCP Server
MCP AI & Machine Learning
Pending Medium
Official Weights & Biases MCP server for querying runs, traces, reports, projects, and documentation through natural language.
MIT wandb/wandb-mcp-server ★ 53
Aescut
A toolbox
A toolbox
your agent can trust
250 skills and MCP servers, vetted for security, permissions, and supply chain trust. Your agent discovers, installs, and checks them at runtime — no manual setup.
Every entry is assessed for permissions, data access, and supply chain risk before it reaches your agent.
One install, full access
Give any agent the
Give any agent the
entire registry
One MCP server. Every skill lookup, risk check, and permission audit — available to Claude Code, Cursor, Windsurf, Codex, or any MCP-compatible agent. No per-tool configuration.
lookup_skill search_skills check_risk list_categories
One command
npx -y @aeptus/aescutWorks with Claude Code, Cursor, Windsurf, and any MCP-compatible agent.
Or add manually
"aescut": {
"command": "npx",
"args": ["-y", "@aeptus/aescut", "--skip-install"]
}Recently Added
Latest skills submitted for security review
Pinecone Assistant MCP Server
MCP AI & Machine Learning
Pending Medium
Official Pinecone Assistant MCP server for retrieving information from Pinecone Assistant knowledge bases.
MIT pinecone-io/assistant-mcp ★ 43
Pinecone Developer MCP Server
MCP AI & Machine Learning
Pending Medium
Official Pinecone developer MCP server for documentation search, index management, record upsert, and search workflows.
Apache-2.0 pinecone-io/pinecone-mcp ★ 65
1Password CLI LLM Docs
Skill Security & Compliance
Pending Low
Official 1Password machine-readable CLI documentation for LLM-assisted workflows and AI coding tools.
Build on 1Password with LLMs
Skill Security & Compliance
Pending Low
Official 1Password docs-backed guide for building LLM and agent workflows with 1Password developer tools.
Agent Skills for CockroachDB
Skill Data & Analytics
Pending Low
Official CockroachDB docs-backed catalog of installable agent skills for database operations and development workflows.
Registry at a glance
What's in the registry
140
Skills
110
MCP Servers
57
Maintainers
13
Categories
Low risk: 125
Medium risk: 100
High risk: 19
Critical risk: 6
Browse by topic → → → → → → → → → → → → →
Categories
Blockchain & Web3 (5 skills)
Skills for smart contract security, blockchain auditing, and Web3 development
Security Research (23 skills)
Skills for vulnerability research, fuzzing, static analysis, and security auditing
Mobile Development (14 skills)
Skills for building mobile apps with React Native, Expo, SwiftUI, and native platforms
AI & Machine Learning (21 skills)
Skills for AI model inference, RAG pipelines, image generation, and ML operations
Payments & Commerce (5 skills)
Skills for payment processing, billing, subscriptions, and financial operations
Productivity & Collaboration (10 skills)
Skills for team communication, project management, and productivity workflows
Security & Compliance (17 skills)
Skills for security monitoring, audit logging, compliance checks, and IAM
Data & Analytics (25 skills)
Skills for querying databases, analyzing data, and generating analytics reports
Cloud & Infrastructure (35 skills)
Skills for managing cloud resources, containers, and infrastructure-as-code
Developer Tools (73 skills)
Skills for building software, testing applications, and creating developer infrastructure
Communication & Writing (3 skills)
Skills for writing documentation, organizational communications, and collaborative authoring
Document Creation (4 skills)
Skills for creating and manipulating office documents (Word, PDF, PowerPoint, Excel)
Design & Creative (15 skills)
Skills for visual design, generative art, branding, and creative asset production
Why this exists
Skills are the biggest unvetted attack vector in AI dev
Every skill and MCP server runs with your agent's permissions — reading files, making requests, executing commands. A single compromised tool can exfiltrate credentials or inject code, and your agent will do it because you told it to install the tool.
We built Aescut internally at Aeptus because we were tired of manually checking every tool. Now it's open source, because safety shouldn't be proprietary.
Read the full story → Open Source
Community Driven
Every vetted skill is published to our public registry as a review card. Submit new skills, contribute reviews, and help the community find safe AI tools.
The registry, review criteria, and MCP server are all open source. Contributions welcome on GitHub.