MCP servers are flaky, insecure or bloat the context

MCP pollutes agent context even when you are not using it. Use a manually invoked skill instead if you don't want to be wasting tokens on every turn and bloating up agent context making it dumber in the process.

cheema33 · Sep 30, 2026 · comment in “You said no MCP”

It is bizarre that Marco said MCP is hard to compose. It's like he doesn't understand how composability works.

0xbadcafebee · Sep 30, 2026 · comment in “You Said No MCP”

> While a lot of things have improved about MCP, quite a few have not. The biggest issue with MCP continues to be that it’s hard to compose.

andrewingram · Sep 30, 2026 · comment in “Pi.dev: You Said No MCP”

AI tools failing to call CLI tools properly and looping trying to figure out the right arguments/input/output). Now, if you don't care about authentication, don't care about wasted tokens, don't care about the AI screwing up CLI tool calls or needing to be trained on every one, and if you don't need to call something on a server, then MCP is useless to you.

0xbadcafebee · Sep 21, 2026 · comment in “MCP was always a bad idea?”

That doesn't work well for enterprise-managed MCP, where you actually do want the user to overtly authorize their agent to user their identity for MCP services, rather than the MCP server just setting a user ID in an HTTP header somewhere and everyone hoping for the best.

mpyne · Aug 22, 2026 · comment in “New MCP Roadmap”

MCP auth has been a huge pain point for us at C1 - both for our own internal MCP use and in our in-product MCP Gateway - so very glad to see this landing.

russell_h · Jun 19, 2026 · comment in “Zero-Touch OAuth for MCP”

* mcp integration and debugging is dreadful, just a combination of generic "an error ocurred" and sometimes nothing at all * MCP only supports OAuth for shared connectors but auth key doesn't work even with "local" servers that are not necessarily local, just the config is local.

lbriner · Apr 15, 2026 · comment in “Elevated errors on Claude.ai, API, Claude Code”

It was the best way before, and its the best way now. MCP doesn't scale and it has a bloated unnecessarily complicated spec.

noodletheworld · Apr 10, 2026 · comment in “I still prefer MCP over skills”

And what's so powerful about that? Well for one it elegantly solves the main problem most people had with MCP in the first place - context bloat.

isaacrolandson · Mar 17, 2026 · story in “Show HN: MCP Isn't Dead. You're Just Using It Wrong”

MCP is great when the agent decides what to run - it discovers tools without you scripting the interaction. The real problem isn't MCP vs CLI, it's that MCP originally loaded every tool definition into context upfront.

flash_us0101 · Mar 16, 2026 · comment in “Chrome DevTools MCP (2025)”

They're fast when they have what you need, but they don't cover every repo, they go stale, and there are hundreds of questions about any codebase that can't be pre-answered. GitHub MCP checks out files one at a time, which burns through context tokens fast and doesn't scale to large codebases.

instalabsai · Feb 13, 2026 · story in “Show HN: Instagit – MCP server that answers questions about any GitHub repo”

Yeah using Claude-Desktop is broken current, all my mcp tool calls keep failing and retrying, then it'll work for a bit, and stop again.

bastard_op · Feb 3, 2026 · comment in “Anthropic is Down”

Instead of uploading my private contracts to OpenAI, I built ClawRAG – a self-hosted RAG engine that connects to OpenClaw via MCP (Model Context Protocol). Now I can ask "What did the contract say about liability?" and get cited answers, not hallucinations.

2dogsanerd · Feb 1, 2026 · story in “Show HN: Self-hosted RAG with MCP support for OpenClaw”

I built PolyMCP, a framework to make building MCP (Model Context Protocol) servers and integrating agents much easier. Working with MCP and agent tooling often comes with repetitive or frustrating challenges:

justvugg · Jan 23, 2026 · story in “Show HN: PolyMCP – Simplifying MCP Server Development and Agent Integration”

MCP servers can use OIDC to connect to your auth server right now: https://modelcontextprotocol.io/docs/tutorials/security/authorization You give the following abstractions, which I think are interesting thought experiments but unconventional and won't work at all:

skipants · Jan 12, 2026 · comment in “Why Ontario Digital Service couldn't procure '98% safe' LLMs (15M Canadians)”

- Should we integrate with security tools like Invariant Labs' mcp-scan? - Above all, are we solving a real pain point.

belai · Nov 17, 2025 · comment in “Show HN: MCPShark – Wireshark for Model Context Protocol Traffic”