Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/rodrigopg/whatsapp-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/rodrigopg/whatsapp-mcp/clone)<a href="https://agentmods.dev/commands/rodrigopg/whatsapp-mcp/clone"><img src="https://agentmods.dev/badge/commands/rodrigopg/whatsapp-mcp/clone/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/rodrigopg/whatsapp-mcp/clone"><img src="https://agentmods.dev/badge/commands/rodrigopg/whatsapp-mcp/clone.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00029 | $0.00643 |
| Opus 5.5 | $0.00012 | $0.00257 |
| Sonnet 5.5 | $0.00006 | $0.00129 |
| Haiku 4.5 | $0.00003 | $0.00064 |
Grade A, and why
clone scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 39 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Turn the user's sent voice notes into a dataset for voice cloning or TTS fine-tuning (e.g. Qwen3-TTS). Everything stays on this machine; nothing is uploaded.
Hard rules
- Only the user's own voice: the script reads
is_from_me = 1audio only. Never add audios received from other people — that is someone else's voice and needs their consent. - The dataset holds private conversation text (names, health, money). Tell the user not to publish it or push it to a public repo/Hub.
1. Pick the account and the store
If more than one whatsapp* account exists, ask which one holds the user's voice notes (default: the one with more sent audios). The store is the bridge's store/ dir (has messages.db and one folder per chat), e.g. ~/whatsapp-mcp/whatsapp-bridge/store on the VPS or whatsapp-bridge/store in a local checkout.
2. Build
cd whatsapp-mcp-server
uv run --with imageio-ffmpeg python clone_voice.py --store <store> --out ~/voice-dataset --max-sec 25
Options: --since 2025-01-01 (recent voice = better match to current voice), --chat <jid>, --limit N, --min-sec. imageio-ffmpeg brings a static ffmpeg; set FFMPEG=/path to use another.
The script keeps clips of 2-25 s whose transcript already exists (the bridge's transcription sweep) and whose characters-per-second is plausible, converts them to 24 kHz mono 16-bit wav, and writes:
wavs/*.wavtrain_raw.jsonl—{"audio", "text", "ref_audio"}per line (the Qwen3-TTS fine-tuning input;ref_audiois one 5-10 s clip shared by all rows)metadata.csv— LJSpeech styleid|textreport.txt— clip count, hours, skipped counts
Audios not yet downloaded to disk are skipped (no_file); download more with download_media and re-run if the count is too low.
3. Report and hand off
Tell the user: clips, hours, the output path, and the skipped breakdown. Rough guide: ~1 h of clean speech is enough for a clone; 3 h+ is plenty for fine-tuning.
Next step for Qwen3-TTS: run its finetuning/ data-prep on train_raw.jsonl (adds audio codes), then the SFT script — follow that README. Spot-check 5 clips by ear first: transcripts come from automatic transcription, so a few will be wrong; bad text hurts more than fewer clips.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 39 lines · 29 tokens per session scan A 7b6b7d5a570f
clone is a command published in the GitHub repository rodrigopg/whatsapp-mcp (4 stars, last pushed today), licensed MIT. It adds 29 tokens to every session and 643 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-16.
Other commands, from other repositories
setup-video-vision
Interactive setup wizard for claude-video-vision — configure backend, whisper, frames, and verify dependencies.
typegen
Run Sanity TypeGen and troubleshoot type generation issues.
review-epo-claims
Analyze patent claims for EPO Art. 84 EPC compliance - clarity, conciseness, support by description.
full-review
Complete patent application review (claims + specification + formalities in parallel).
review-formalities
Check patent application formalities (abstract, title, drawings per MPEP 608).
autosearch
Deep research via AutoSearch v2 tool-supplier — clarify → select channels → search → synthesize.