The 18,000-Token Context Tax Agents Are Finally Refusing to Pay

When Mario Zechner put a meter on the popular browser-automation MCP servers, the readings were hard to believe: Chrome DevTools MCP needs 26 tool definitions, roughly 18,000 tokens — about 9% of a 200,000-token context window — before the agent has done anything useful. Playwright MCP was only slightly lighter at 21 tools and 13,700 tokens. His conclusion in a November 2025 post was close to heretical for its time: what if you don't need MCP at all? He wired browser control into his own coding agent, Pi, using a handful of Bash scripts and a 225-token README, and left the protocol out for nearly a year.

MCP won the protocol war anyway — it is now built into nearly every developer-facing agent — but look at how the holdouts are finally adopting it, and you get the real story. Pi 1.0 shipped MCP support on October 1, and the manner of adoption is the interesting part. Tool definitions do not go into the prompt. Each connected server gets a one-line description, and the agent finds tools on demand through Codemode, a QuickJS sandbox where scripts call tools, run operations concurrently, and process results before anything reaches the model. That kills Zechner's other complaint: MCP results previously had to round-trip through the agent's context just to be saved to disk or combined with other output. Per-tool exposure controls let a setup expose search_code directly, keep get_* behind Codemode, and block delete_* outright. Codemode carries a 3,000-token default budget for tool declarations, and the Pi 1.0 release notes claim a leaner Codemode with roughly 40% fewer prompt tokens — a GPT-5.6 request fell from about 5,300 to 3,300 tokens.

Source article image
Source image 1

The pattern is bigger than one coding agent. The protocol survived; the naive integration — dump every JSON schema into the system prompt and hope the model figures it out — is quietly dying. Lazy tool discovery, progressive disclosure, and sandboxed execution are the same tricks front-end developers have used for years to keep pages lean, now applied to agent tooling, because context is where an agent's attention lives. Eighteen thousand tokens of tool descriptions is not just an invoice problem; it is noise competing with your actual task, and anyone who has watched an agent pick the wrong tool out of forty has already paid that tax. Multiply it across a dozen servers and a team running agents all day, and context bloat becomes a line item on the cloud bill. Worth asking: if you audited your own agent setup, which MCP servers would actually survive a token-cost review?

Source article image
Source image 2

Sources

Comments

Post a Comment

Popular posts from this blog

AI Is Starting to Feel Less Like a Gadget and More Like Infrastructure

Image
Modern AI doesn't replace infrastructure — it runs on top of it. (Photo: Unsplash) For a while, AI coverage had the energy of a mall demo kiosk: very shiny, slightly chaotic, and usually one prompt away from embarrassment. Lately, though, the more interesting story is less about flashy demos and more about infrastructure. A useful CNBC piece argued that AI is not simply going to vaporize enterprise software overnight, and that sounds about right. In real companies, software does not disappear just because a model can summarize a meeting or write a decent SQL query. What actually happens is messier and more practical: AI starts sneaking into the plumbing. It shows up in search, support workflows, procurement tools, analytics dashboards, and data pipelines. The real winners may not be the loudest chatbot wrappers, but the vendors that make enterprise systems easier to operate, easier to query, and slightly less soul-crushing for the humans trapped inside them. That is less cinematic ...
Read more

When Two AI Bots Finally Learned to Talk in Discord

I spent part of the day doing something that sounds like the setup for a bad joke: getting two local AI assistants to talk to each other in the same Discord channel. Not through a web UI. Not by bouncing prompts manually between windows like a human message queue. In the actual shared channel, where both could see the conversation and react to each other. The funny part is that the problem was not intelligence. It was manners. Both bots were defensive by default around bot-authored messages, which is the sensible setting if you do not want your infrastructure turning into a recursive support group. The downside is obvious: if both assistants treat other bots as suspicious background noise, they will never coordinate on anything more useful than silence. The fix was to make both sides accept bot-authored messages only when they were explicitly mentioned. That detail matters more than it sounds. Blanket bot acceptance is how you end up with two enthusiastic systems discovering each oth...
Read more

Enterprise AI Is Quietly Becoming a Systems Integration Problem

The funniest thing about enterprise AI in 2026 is that the flashy part is basically over. The demos are still shiny, sure, but the real story now looks a lot less like science fiction and a lot more like infrastructure planning, governance, and somebody in IT asking who exactly is paying for all these agents. Over the past few weeks, Microsoft has been pitching Agent 365 and a bundled Microsoft 365 E7 “Frontier Suite” aimed at governing and securing fleets of workplace agents, Google has introduced a new Workspace add-on for higher AI usage tiers, and Anthropic has thrown $100 million behind a partner network to help companies move Claude deployments from pilot mode into something that can survive contact with procurement, compliance, and existing systems. That cluster of announcements says something pretty clear: enterprise AI is no longer mainly a model race. It is becoming a packaging, controls, and implementation race. The clever model still matters, obviously, but once every vendo...
Read more