--- title: Does Your MCP Server Fill the Context Window? We Measured description: Codex peaked at 18.1% of its window, Claude Code at 4.1%, yet both billed ~210k input tokens per question. What fills the context window, and what MCP adds. canonical: https://spiritroom.app/blog/mcp-context-window-usage/ published: 2026-10-21 updated: 2026-10-21 --- > Markdown version of https://spiritroom.app/blog/mcp-context-window-usage/ for agents. Index: /llms.txt [Blog](/blog/) / [How agents load MCP tools](/guides/how-agents-load-mcp-tools/) # Why your MCP server fills the context window, and which agents actually pay for it “My MCP server uses too many tokens” can mean two different things: the window gets full, or the bill gets big. They have different causes and different fixes. We measured both on real runs. Leonid Bugaev · Published 21 Oct 2026 **TL;DR** - **Two numbers:** peak (the largest single request: how full the window gets) and cumulative (all requests summed: what's billed). - **Peak was fine:** Codex averaged 18.1% of its 258,400-token window, Claude Code 4.1% of 1M. - **Cumulative was the cost:** about 218k (Codex) and 210k (Claude Code) input tokens per simple read question. - **MCP results were 12–46%** of the largest request. Tool definitions, search output and the app's own prompt were often bigger. - **Preloading apps hurt most:** a raw 99-tool set took OpenCode to 51% of its window on average (max 84.5%). [Two numbers, not one](#two)[What we measured live](#live)[What fills the window](#what)[Which agents pay](#who)[What to cut first](#fix)[How we measured](#how-we-measured)[FAQ](#faq) ## Two numbers, not one - **Peak**: the largest single model request in a run. That's how full the window got. If it overflows, the app truncates or compacts. - **Cumulative**: every request in the run added up. Agents make several requests per question, each resending the conversation so far. That's what's billed (mostly as cached input). A server can be safe on peak and expensive on cumulative. That's what we found. ## What we measured live Codex CLI and Claude Code answered the same 8 read-only jobs, 5 runs each, on a 99-tool production MCP server. Codex: avg peak, % of 258,400 *46,794 tokens (18.1%)* Claude Code: avg peak, % of 1M *40,910 tokens (4.1%)* Codex: avg cumulative per question *218,032 tokens* Claude Code: avg cumulative per question *209,722 tokens* No live run came close to overflowing; Codex's worst single request was 24.4% of its window. But a simple read question cost 114k–301k input tokens on Codex, depending on the job, because it made 4–8 model requests. **Find out what your MCP server costs each agent.** We measure peak and cumulative context per job and per agent app on your server, and show which tools and results to cut first. [Measure your MCP →](/#contact) ## What fills the window In the live Codex runs, the average largest request (46,794 tokens) split roughly as: App prompt + hidden *17,062* Tool search output *16,366* MCP results *11,734 (23%)* Across all our groups, MCP results were **12–46%** of the largest request. Clear-cut waste inside them was small (0.2% of the largest request in live runs) except where a tool returned its own audit log: 7.1%. Records nobody used were the bigger, fuzzier share: up to 12.5%. More: [reduce tool response size](/blog/reduce-mcp-tool-response-size/). **Bytes aren't tokens.** One overview result was 77,123 bytes on the wire and 14,517 tokens. What the model sees can also be less than the server sent: some apps cut long results. See [output limits by agent](/blog/mcp-tool-output-limits-by-agent/). ## Which agents pay | App | How it loads 99 tools | Effect | | --- | --- | --- | | Claude Code (tool search on) | Names only, definitions on demand | +1.7k tokens | | Codex | Tool search | +0 up front; search output 11–16k per request | | OpenCode | Preloads every definition | +73k tokens; avg peak 51% of 204.8k, max 84.5% | With a trimmed tool set (v2), OpenCode's largest request shrank 4–5×, to 11.2% of its window. Details: [how much context MCP tools use](/blog/how-many-mcp-tools-is-too-many/). **Go deeper:** [How much MCP output is waste](/blog/mcp-context-pollution/), [Tool count and context](/blog/mcp-task-token-cost-claude-code-vs-codex/), [How Codex finds MCP tools](/blog/how-codex-finds-mcp-tools/). ## What to cut first 1. **If users run preloading apps:** trim definitions. They're resent in every request. 2. **If cumulative is high:** cut the number of requests. Fewer round trips means fewer resends. Make the obvious tool return the answer in one call. 3. **Then results:** drop audit logs and records the question didn't ask for; add filters and summaries. 4. **Measure tokens** per job and per app, peak and cumulative, not bytes. Back to [the MCP tool discovery guide](/guides/how-agents-load-mcp-tools/). ## How we measured - **Live:** Codex CLI 0.161.0 and Claude Code 2.1.301, 8 read-only jobs × 5 runs, 16 Oct 2026, through a read-only proxy. Peak = largest request's input tokens from the app's session log; cumulative = sum over the run. - **Breakdown:** o200k\_base token counts for each part of the largest request; the residual is app-reported input minus attributed tokens (±3% from mixing tokenizers). - **Mock:** 150 runs on a stub of the same 99 tools for the OpenCode and v2 figures; captured traffic for the definition overhead. - **Limits:** one workspace, read-heavy jobs, one day, single model versions. Windows are as each app reports them. ## FAQ ### Why does my MCP server use so much context? Three things add up: tool definitions (or tool search output), tool results, and repeated requests. In our runs MCP results were 12–46% of the largest request; the rest was definitions, search output, the app's prompt and the model's own turns. ### What's the difference between peak and cumulative context? Peak is the largest single request in a run, which decides whether the window overflows. Cumulative is the sum over every request in the run, which is roughly what you pay for. An agent making 6 requests of 40k tokens peaks at 40k but bills 240k. ### Which agent apps load all MCP tools up front? In our tests OpenCode preloaded all tool definitions (+73k tokens for 99 tools). Claude Code with tool search on added about 1.7k; Codex loads tools through search. ### Are bytes on the wire the same as tokens? No. One 77,123-byte result was 14,517 tokens. Measure tokens with a tokenizer, not response size. **Find out what your MCP server costs each agent.** We measure peak and cumulative context per job and per agent app on your server, and show which tools and results to cut first. [Measure your MCP →](/#contact) ## Keep reading - Up[How agent apps load MCP tools: what the model sees (guide)](/guides/how-agents-load-mcp-tools/) - Related[How many MCP tools is too many? Real token costs across five agents (2026)](/blog/how-many-mcp-tools-is-too-many/) - Related[How to reduce MCP tool response size: measured before and after](/blog/reduce-mcp-tool-response-size/) - Related[How much of your MCP output is waste? Measuring context pollution in agent runs](/blog/mcp-context-pollution/) - Related[Same job, two agents: what one MCP task really costs in Claude Code and Codex](/blog/mcp-task-token-cost-claude-code-vs-codex/)