Skip to content

API Call Budget

Tencent Docs MCP rate limits: 100/day (free) · 1000/day (member) · 2000/day (plus) · 20,000/month (all tiers).

Every tools/call counts. tools/list does not.

Per-command cost

CommandWithout cacheWith cacheNotes
qqdocs ls10 (if <5 min old)listRecent · SWR TTL 300s
qqdocs ls --dates1 + N0 (if <5 min old)+query_file_info per doc
qqdocs ls <folder>1 + 10 (if <5 min old)folder_list + resolveFolderId
qqdocs sync2listRecent + folder_list; manual only
qqdocs read <ref>1read_doc
qqdocs delete <ref> (dry-run)2getDocInfo + readDoc; writes doc-cache
qqdocs delete <ref> --confirm00reads doc-cache (5 min TTL); skips both calls
qqdocs info <ref>1query_file_info
qqdocs rename <ref>1rename_doc
qqdocs cp <ref>1copy_doc
qqdocs open <ref>0browser open, no API
qqdocs create1–2create_doc + optional set_permission
qqdocs import3–5upload + start + poll until done
qqdocs perm get1query_permission
qqdocs perm set1set_permission
qqdocs space list1list_spaces
qqdocs space ls1list_space_nodes
qqdocs space mkdir/mkdoc/link1one node creation call
qqdocs space rm1delete_space_node
qqdocs space move1move_doc_to_space
qqdocs canvas read1 per pageread_smart_canvas; --all multiplies
qqdocs canvas find1+read + search
qqdocs canvas edit1edit_smart_canvas
qqdocs usage0local file only
qqdocs usage calibrate0local file only
qqdocs tools0tools/list (not rate-limited)
qqdocs raw1one arbitrary tool call

Caching layers

1. SWR list cache — ~/.qqdocs/cache.json

  • Written by qqdocs sync and refreshed by qqdocs ls (background fetch)
  • TTL: 300 seconds — repeated ls within 5 min costs 0 calls
  • Contains: file_id, title, url, mtime, owner per entry
  • Used by: qqdocs ls, tab-completion candidate resolution

2. Doc content cache — ~/.qqdocs/doc-cache.json

  • Written by qqdocs delete <ref> dry-run (after fetching content + info)
  • TTL: 5 minutes — safe window for the two-step delete confirm flow
  • Eliminates: 2 API calls (getDocInfo + readDoc) on delete --confirm
  • Stale entries are pruned on next write

3. Usage tracker — ~/.qqdocs/usage.json

  • Incremented locally after every tools/call
  • No API calls; used for quota progress display and tier inference

High-cost patterns to avoid

  • qqdocs ls --dates in a loop — N extra query_file_info calls per run
  • Scripting qqdocs delete without reusing the same session (TTL is 5 min)
  • qqdocs sync in a cron job at high frequency — costs 2 calls each time
  • qqdocs canvas read --all on large documents — one call per page

Checking usage

bash
qqdocs usage                          # show progress bars
qqdocs usage calibrate --today 45     # correct today's count
qqdocs usage calibrate --tier member  # set tier after upgrade