tokenwatch.wyrdwerk.com

Last change

None yetUnchanged since first read.

Declared read-only

9 of 21

llms.txt

Present

1. about_tokenwatchReads only

   Read-only. Returns the TokenWatch WebMCP operating brief (rules and page capability map) plus the URL of the full skill. Call this before ranking, filtering, or comparing so you use {provider, id} identity and the live table correctly.

   First seen 11 Oct 2026. Unchanged since.

2. apply_presetCan change things

   Applies a named mix preset (agentic, balanced, heavy-output, no-cache) and re-renders the table. Does not change filters. For operational details, call about_tokenwatch.

   First seen 11 Oct 2026. Unchanged since.

3. clear_filtersCan change things

   Resets search, ZDR, subscription, promo, group-by, min-intelligence/coding/agentic, benchmarked-only, hide-batch (back to on), cache-only, max blended, min tok/s, and HQ filters. Keeps the current workload. Re-renders the table. For operational details, call about_tokenwatch.

   First seen 11 Oct 2026. Unchanged since.

4. compare_modelsCan change things

   Add, remove, clear, or set the compare tray (max 6). Identity is {provider, id} from get_view, never a rank. Optionally open the side-by-side compare modal. Re-renders the table and tray. For operational details, call about_tokenwatch.

   First seen 11 Oct 2026. Unchanged since.

5. download_cost_cardCan change things

   Triggers a PNG download of the cost card for one offering in the current view. In-app browsers may block the download; use get_share_url as a fallback. For operational details, call about_tokenwatch.

   First seen 11 Oct 2026. Unchanged since.

6. explain_rankingReads only

   Read-only deterministic explanation of why the current #1 ranks ahead of #2 using the table's active sort and direction (cost, blended rate, speed, prices, context, provider, org, or model name). Includes the active ranking metric, ranking values, cost components, and how many offerings were excluded because they cannot serve the requested mix. Does not change the table. For operational details, c

   First seen 11 Oct 2026. Unchanged since.

7. export_csvCan change things

   Triggers a CSV download of the current ranked results (same as the Export CSV button). In-app browsers may block the download; the result still reports filename and rowCount. Prefer get_share_url if the file does not appear. For operational details, call about_tokenwatch.

   First seen 11 Oct 2026. Unchanged since.

8. get_catalog_infoReads only

   Read-only. Page name, pricing.json generated_at, and catalog size. Use this instead of inventing how fresh the data is. For operational details, call about_tokenwatch.

   First seen 11 Oct 2026. Unchanged since.

9. get_modelReads only

   Read-only detail for one offering in the current view: pricing, context, cache, ZDR/subscription, benchmarks, Neuralwatt energy if present, capabilities, speedP50, speedReported, speedWindow, ttftP50 (seconds), and ttftWindow. Provider-reported decode speed is not a median. Requires {provider, id} from get_view. For operational details, call about_tokenwatch.

   First seen 11 Oct 2026. Unchanged since.

10. get_share_urlReads only

    Read-only. Returns the current page URL including the hash of mix/filters so the human can open the same view in any browser without ChatGPT. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

11. get_viewReads only

    Read-only snapshot of the TokenWatch text calculator the human is looking at: current mix, modes, filters, active sort, compare tray, rowCount, and the top ranked offerings (rank, provider, id, name, cost, blended $/M, quantization, zdr, speedP50, speedReported, speedWindow, ttftP50 in seconds, ttftWindow, intelligence, coding, agentic). Provider-reported decode speed is speedReported, not a media

    First seen 11 Oct 2026. Unchanged since.

12. highlight_tradeoffCan change things

    Fills the compare tray with cheapest / fastest / ZDR-cheapest / smartest (highest AA intelligence with a score) from the current view and opens the compare modal. Re-renders the table. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

13. list_presetsReads only

    Read-only list of named mix presets (agentic, balanced, heavy-output, no-cache) with their token mix. Use apply_preset to apply one — that re-renders the table. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

14. open_detailCan change things

    Opens the detail modal for one offering that is in the current view (pricing, benchmarks, Neuralwatt energy). Requires {provider, id} from get_view. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

15. recommend_modelReads only

    Read-only workload recommendation using TokenWatch's shared engine. The default response is compact: three fully explained picks plus compact candidate groups with total counts. For chat-assistant and creative-writing, also read the Arena preference favorite and source attribution/date/license, plus its rating and providers; chat picks remain capability-ranked while creative writing is preference-

    First seen 11 Oct 2026. Unchanged since.

16. recommend_providerReads only

    Read-only provider ranking for a canonical model and workload using the shared engine. The default response fully explains the top three providers and returns the rest as compact rows with total counts. Confirmed capabilities and context are hard gates; unknown required metadata is never treated as confirmed. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

17. set_filtersCan change things

    Partial update of filters: provider search, model search, ZDR only, subscription only, promos only, group-by, min intelligence/coding/agentic, benchmarked only, hideBatch (default true), cacheOnly, maxBlended $/M, minToks, hq country. Re-renders the table. Workload is unchanged. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

18. set_sortCan change things

    Sets the active table sort and direction, then re-renders the same results. Sortable columns: org, provider, model, quantization (best effort), input, output, cache_read, context, speed, ttft, intelligence, coding, agentic, blended, and cost. Text fields sort alphabetically; numeric fields sort ascending or descending. ttft is time-to-first-token in seconds (lower is faster). intelligence/coding/a

    First seen 11 Oct 2026. Unchanged since.

19. set_workloadCan change things

    Sets tokens or budget, input/cache/output mix, per-session vs monthly, and forward vs budget mode. Re-renders the results table. Mix percentages must sum to 100 (±0.5); they are not silently renormalized. Partial updates are allowed. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

20. snapshot_compareCan change things

    Opens the compare modal (needs ≥2 selected) and triggers a PNG download of the comparison card. In-app browsers may block the download; use get_share_url as a fallback. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

21. switch_catalogCan change things

    Navigates to another TokenWatch catalog (text, image, video, benchmarks). The next page may register a thinner tool set. This leaves the current page. For operational details, call about_tokenwatch.

    First seen 11 Oct 2026. Unchanged since.

Run this site? This page shows what your site told our crawler, ToolChronicleBot. To be left out, disallow it in robots.txt or write to hello@toolchronicle.com.