IdeaWell

Project ideas from Hacker News discussions.

I still prefer MCP over skills

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

4Dominant Themes from the HN Thread

# Theme Supporting Quote
1 Skills excel at encoding non‑repeatable, intuitive knowledge – they can bundle code and be stored as markdown files. “Skills are good for instilling non‑repeatable, yet intuitive or institutional knowledge.” – robotobos
2 MCP provides a standardized, portable, and secure way to expose external tools, especially in enterprise settings – it defines auth and tool signatures that multiple agents can share. “MCP makes a lot of sense for enterprise IMO. It defines auth and interfaces in a way that’s a natural extension of APIs.” – bicx
3 Tool‑discovery and context‑bloat matter – loading whole skill files inflates the LLM’s context, whereas MCP can be discovered lazily (progressive disclosure). “Using a skill often requires loading the entire SKILL.md into the LLM’s context window, rather than just exposing the single tool signature it needs.” – contextbloat
4 Sandboxing and secret handling favour MCP for shared or remote environments – agents in sandboxes or on locked‑down devices need a controlled API surface rather than raw CLI access. “There is no world in which an enterprise is not OK with an agent having access to a CLI but is OK with possibly getting prompt injected from a random web search.” – woeirua

Bottom line: Skills are great for local, knowledge‑driven workflows, while MCP offers a reusable, secure, and discoverable bridge to external services—especially when multiple users, sandboxes, or enterprise policies are involved. Both can coexist; the choice depends on where the work happens and who needs to control the access.

🚀 Project Ideas

Progressive MCP Dispatcher#Summary

  • Provides lazy, context‑aware discovery of MCP tools to eliminate context‑window bloat and let large agents scale efficiently.

Details

Key Value
Target Audience AI developers, hobbyist agents, and teams building context‑intensive assistants
Core Feature On‑demand concise tool signatures; dynamic loading/unloading of MCP descriptors
Tech Stack Node.js/TypeScript, MCP SDK, React config UI, SQLite cache
Difficulty Medium
Monetization Hobby

Notes- Why HN commenters would love it (quote users if possible): “MCP has severe context bloat” – this directly solves that pain.

  • Potential for discussion or practical utility: Enables agents to handle dozens of tools without ever hitting token limits; can be packaged as a reusable library for any MCP‑aware agent framework.

Vaulted MCP Auth Gateway

Summary

  • Injects per‑user OAuth tokens into MCP calls while keeping secrets hidden from the agent, solving enterprise security concerns.

Details

Key Value
Target Audience Enterprise security officers and developers of internal agents
Core Feature Secret‑aware MCP proxy that swaps scoped tokens at runtime and masks them from downstream agents
Tech Stack Python FastAPI, HashiCorp Vault, OAuth2 flows, MCP server wrapper
Difficulty High
Monetization Revenue-ready: Enterprise SaaS subscription per active user

Notes- Why HN commenters would love it (quote users if possible): “In an organisation we can’t limit MCP access. It’s all or nothing.” – this adds granular limits.

  • Potential for discussion or practical utility: Provides a clean RBAC layer for remote MCPs, making compliance audits straightforward and enabling safe multi‑tenant AI services.

Skill‑MCP Generator

Summary

  • Automatically converts MCP tool specifications into ready‑to‑use markdown skill files, keeping documentation in sync with the underlying protocol.

Details| Key | Value |

|-----|-------| | Target Audience | AI engineers and maintainers of MCP‑heavy workflows | | Core Feature | Schema‑to‑markdown templating that outputs Skills with usage examples and auto‑generated help sections | | Tech Stack | Go, JSON schema parser, markdown templating engine | | Difficulty | Low | | Monetization | Hobby |

Notes

  • Why HN commenters would love it (quote users if possible): “Skills are markdown files, so they definitely can’t call APIs.” This bridges that gap.
  • Potential for discussion or practical utility: Eliminates duplicated documentation effort; any change to an MCP server instantly reflects in the corresponding skill, reducing maintenance overhead.

Enterprise MCP Permission Broker

Summary

  • Enforces fine‑grained RBAC policies on remote MCP calls, making MCP adoption safe for regulated enterprises.

Details

Key Value
Target Audience Large organizations, compliance teams, and internal platform engineers
Core Feature Identity‑driven policy engine that maps user roles to allowed MCP tool scopes and actions
Tech Stack .NET Core, LDAP/Active Directory integration, Policy Decision Point (PDP) middleware, MCP server plug‑in
Difficulty High
Monetization Revenue-ready: Per‑seat annual license with tiered feature packs

Notes

  • Why HN commenters would love it (quote users if possible): “MCP is better for restricted environments because you know exactly what it can or cannot do.” This makes that control explicit and enforceable.
  • Potential for discussion or practical utility: Turns MCP into a governed marketplace, encouraging enterprise adoption while satisfying security audits; can be sold as a managed service or on‑prem license.

MCP Workflow Studio#Summary

  • Visual low‑code interface for building reusable agent workflows that chain MCP tools and CLI scripts without writing code.

Details

Key Value
Target Audience Non‑technical power users, citizen developers, and product teams seeking rapid automation
Core Feature Drag‑and‑drop workflow canvas that assembles MCP endpoints, CLI wrappers, and prompt templates into executable agent tasks
Tech Stack React front‑end, Node.js backend, WebAssembly sandbox for CLI execution, SQLite persistence
Difficulty Medium
Monetization Revenue-ready: Tiered subscription (Starter, Pro, Enterprise) with usage‑based pricing

Notes

  • Why HN commenters would love it (quote users if possible): “The biggest value of MCP is that it’s an API abstraction.” This abstraction becomes a visual, reusable asset for non‑devs.
  • Potential for discussion or practical utility: Lowers the barrier to AI‑driven automation; can become a marketplace of shared workflows, driving community growth and network effects.

Read Later