August 10, 2026
Token drama goes full cartoon
Show HN: Mcptoon – MCP CLI client that cuts tool discovery tokens by 97%
A tiny app promises huge savings, but the comments instantly turned into a trust war
TLDR: mcptoon claims it can dramatically shrink the machine-readable clutter AI tools send back and forth, freeing more room for real tasks. Commenters were split between impressed curiosity and full-blown suspicion, with critics questioning the math, the examples, and whether this is just old-school compression in flashy packaging.
A new Show HN project called mcptoon is pitching a very online dream: make chatbots stop wasting so much space on bloated machine chatter and save up to 97% of that overhead. In plain English, it’s a tiny command-line helper that swaps bulky formatting for a stripped-down version, so AI assistants can spend more room on actual work instead of digital filler. The creator sells it as small, fast, and usable across basically every major AI coding tool.
But the real entertainment was in the comment section, where readers went from curious to combative in record time. One camp was genuinely intrigued, with one builder saying they manage nearly 100 tools in a video editor and wanted to learn the trick. The other camp came in swinging. Critics accused the post of sounding like a late-night infomercial, questioned whether the examples were even comparing the same information, and zoomed in on tiny details like replacing “null” and line breaks with weird symbols. That kicked off the classic internet brawl: is this clever efficiency, or just compression wearing a fake mustache?
The funniest energy came from the sheer skepticism. One commenter basically rolled their eyes at the whole “97%” pitch, calling out “AI hustlers” and blaming messy tool design instead. So yes, mcptoon may be trying to cut token waste — but the comments were busy spending absolutely zero words on politeness.
Key Points
- •The article presents mcptoon as an MCP CLI client that outputs a compact format called TOON instead of JSON to reduce token usage.
- •It claims MCP tool discovery across multiple servers can consume around 10,000 tokens and total protocol overhead can reach 40,000-70,000 tokens in a 128K context window.
- •The article reports token savings of about 97% for tool discovery, 56% for structured tool results, and 10% for raw HTML or text results.
- •mcptoon is described as a zero-dependency, pure Python tool of roughly 50KB that supports Python 3.10+ on Windows, macOS, and Linux.
- •The post says mcptoon works with shell-command-capable AI agents including Claude Code, Codex, OpenCode, Cursor, and CatPaw, using one shared config file.