50,268 tokens across 41 tools — very heavy (≥ 30K). Measured 2026-09-30 under methodology v1.0.
An Anthropic request carries 49,877 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) | comfyui-mcp v0.52.203 |
| status | measured |
| tokenizer | tiktoken / o200k_base |
| launch command | npx -y comfyui-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 | c15058cd28d060963b537e7c4d10895d56c2730ea0382b1a037d44e16cd8bb1b |
| category | community |
| source | https://github.com/artokun/comfyui-mcp |
| tool | tokens | share | description | input schema |
|---|---|---|---|---|
| download_model | 3,986 | 7.9% | 2,111 | 1,823 |
| generate_image | 3,899 | 7.8% | 1,278 | 2,586 |
| node_pack | 2,961 | 5.9% | 1,401 | 1,521 |
| get_image | 2,642 | 5.3% | 1,151 | 1,462 |
| get_workflow | 2,521 | 5.0% | 1,124 | 1,365 |
| install_comfyui | 2,501 | 5.0% | 1,412 | 1,035 |
| list_packs | 2,314 | 4.6% | 1,681 | 595 |
| comfy_cli | 2,032 | 4.0% | 793 | 1,204 |
| train_start | 1,736 | 3.5% | 683 | 1,023 |
| list_local_models | 1,731 | 3.4% | 978 | 726 |
| install_custom_node | 1,614 | 3.2% | 837 | 747 |
| runpod | 1,570 | 3.1% | 960 | 562 |
| enqueue_workflow | 1,536 | 3.1% | 835 | 670 |
| train_prepare_dataset | 1,516 | 3.0% | 781 | 704 |
| create_workflow | 1,478 | 2.9% | 380 | 1,075 |
| upload_image | 1,431 | 2.8% | 613 | 792 |
| queue | 1,351 | 2.7% | 851 | 471 |
| get_defaults | 1,102 | 2.2% | 643 | 433 |
| get_history | 1,040 | 2.1% | 596 | 417 |
| model_metadata | 1,022 | 2.0% | 742 | 258 |
| apps | 898 | 1.8% | 360 | 513 |
| save_workflow | 808 | 1.6% | 511 | 274 |
| visualize_workflow | 804 | 1.6% | 258 | 519 |
| batch | 742 | 1.5% | 350 | 365 |
| calculate | 656 | 1.3% | 405 | 215 |
| get_system_stats | 625 | 1.2% | 336 | 267 |
| apply_manifest | 625 | 1.2% | 237 | 370 |
| kitchen | 623 | 1.2% | 379 | 223 |
| train_doctor | 560 | 1.1% | 330 | 207 |
| list_api_nodes | 557 | 1.1% | 276 | 259 |
11 smaller tools omitted (3,385 tokens combined) — all of them are in the raw capture.
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.
| date | tokens | tools | release | measured in | change |
|---|---|---|---|---|---|
| 2026-09-03 | 50,640 | 41 | not recorded | docker | — |
| 2026-09-04 | 50,640 | 41 | 0.52.189 | docker | no change |
| 2026-09-05 | 50,776 | 41 | 0.52.199 | docker | +136 |
| 2026-09-09 | 50,776 | 41 | 0.52.202 | docker | no change |
| 2026-09-30 | 50,268 | 41 | 0.52.203 | docker | −508 |
Full series: results/history.csv.
npx -y mcp-context-cost verify results/comfyui-mcp/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.