mcp-context-cost

google-surf — context cost

17,185 tokens across 7 tools — heavy (15–30K). Measured 2026-09-30 under methodology v1.0.

An Anthropic request carries 11,434 of those tokens as tool definitions. What Claude makes of them is not published for this server: its Claude count is missing, or was taken against a capture this measurement has since replaced.

   
server (self-reported) google-surf-mcp v1.1.3
status measured
tokenizer tiktoken / o200k_base
launch command npx -y google-surf-mcp
isolation docker · public.ecr.aws/docker/library/node:22-slim · network bridge · linux/amd64 · network enabled for package fetch; clean FS, no host credentials
env vars supplied none
canonical SHA-256 169cfd8e49a251baf2e62bed239afae4ee1999f6fb00a763fa4c66d35b26c515
category community
source https://github.com/HarimxChoi/google-surf-mcp

Where the tokens are

tool tokens share description input schema output schema
project_memory 7,869 45.8% 380 6,088 1,354
project_memory_search 2,742 16.0% 208 1,130 1,354
search_parallel 1,996 11.6% 482 631 833
search 1,873 10.9% 478 612 735
extract 1,272 7.4% 330 511 381
scholar_search 1,004 5.8% 134 302 519
health 427 2.5% 50 24 304

Each tool is tokenized on its own, so the parts do not sum exactly to the whole: the array adds its own brackets and commas, and the tokenizer merges tokens across object boundaries. The badge number is always the count of the whole array, never a sum of parts.

Over time

date tokens tools release measured in change
2026-09-04 10,948 7 1.0.9 docker —
2026-09-05 10,948 7 1.0.9 docker no change
2026-09-09 10,948 7 1.0.9 docker no change
2026-09-30 17,185 7 1.1.3 docker +6,237

Full series: results/history.csv.

Re-derive it

npx -y mcp-context-cost verify results/google-surf/measurement.json

That re-tokenizes the published capture and checks the count and the hash. If it disagrees with the badge, the badge is wrong — open an issue and it gets corrected.

Badge JSON · All servers · Leaderboard · Methodology