[ Idea ]Found 12 Nov · 1 YC · 1 Product Hunt

Canary

See your WhatsApp number's ban risk before you send.

Should you build this?Don't build

Free MIT-licensed baileys-antiban already covers Canary's local pacing and risk scoring; pooled data won't exist at launch.

6 products · 11 sources13 Nov

The problem

Unofficial WhatsApp agent tools can get your number banned, and nobody knows which behaviours cause it.

Meta's official API rules push open-ended personal assistants onto unofficial routes, and ban-rate claims conflict widely, from about 1 in 5 accounts within a year to bans within 2 to 8 weeks. Builders guess at pacing and can lose the number they use for family, clients and one-time codes.

Seen in 1 YC · 1 Product Hunt

The idea

Canary sits between your agent and WhatsApp, showing your number's risk and pacing sends to reduce the chance of a ban.

1Export your agent's logs and check them for risky sending patterns.
2Canary paces outgoing messages and pauses when warning signs appear.
3Opt in to share anonymous outcomes and compare your number with others.

Price: Free: 1 number, local log check, no pooled data. Solo: Rs 499/month (about $6) per number for the live risk meter, pacing governor, pooled benchmarks and failover alerts. Agency: Rs 2,999/month for 10 numbers plus a client-shareable number-health report. Founding members pay Rs 3,999 for a year.

[ The six checks ]Failed at check 3 of 6

1Not built in alreadyPassed

2An AI agent can't do itPassed

3Nothing similar existsFailed here

4People pay for thisNot checked

5Strong scoresNot checked

6Real sourcesNot checked

1 · Not built in alreadyPassed

2 · An AI agent can't do itPassed

3 · Nothing similar exists

Why it failed

Check 3 · Already exists

A free, MIT-licensed, drop-in Baileys middleware (baileys-antiban) already does Canary's core local job: warm-up, human-like pacing, a 0-100 risk score, auto-pause and webhook alerts. Canary's only real wedge is pooled ban-outcome data, and that data won't exist at launch (free tier has none, paid tier depends on opt-in telemetry).

Who paysThe builder or agency owner pays, using a card or UPI. For a solo builder it is a personal-tool expense. For an agency it is a client-delivery cost, justified because a banned client number means lost leads and an angry client.

Who uses itIndie builders and power users running a personal AI agent on their own WhatsApp (Baileys, WAHA, OpenClaw-style setups), and small Indian agencies that run WhatsApp agents for clients on unofficial gateways.

First 10 usersAdarsh already builds WhatsApp personal agents, so he starts with his own number. (1) Post a build-in-public thread on X with a 30-day instrumented log of his agent's number: which patterns raised risk and what he changed. (2) Offer the free Canary check, which reads exported Baileys/WAHA logs locally, to people who reply or complain about bans on X and in Indian indie-hacker and agent-builder communities. (3) DM hackathon organisers and DevRel contacts whose participants ship WhatsApp bots, and offer free beta numbers for their events. (4) Ask GDG on Campus and Takshila members running WhatsApp bots for campus services to try it. (5) Convert the first 10 into a founding-member cohort. Their opt-in telemetry is also what seeds the dataset.

Why not just AIA user's agent can write a rate limiter, but it can't know what actually gets numbers banned. That knowledge only exists in outcome data pooled across many numbers: pacing, new-contact ratio, session age and whether the number later survived. An agent also can't watch itself. The agent is the thing that misbehaves, so the monitor has to sit outside it and be able to throttle it or move the conversation to a fallback channel. A single user can only learn from their own ban, which is the one data point they don't want.

What already exists

6 products checked

ProductLive number risk meterHuman-like pacing governorAnonymous pooled benchmarksFailover and owner alertsShareable number-health reportOverlap with idea

This ideaWhat Canary plans to do

5 / 5

baileys-antiban (kobie3717)Free, drop-in TypeScript library for Baileys that adds warm-up, pacing, a risk meter, auto-pause and alerts. · the reason it failed

[ What would make it pass ]3 steps

01
Seed pooled outcome data from a free beta cohort before charging anyone.
02
Beat baileys-antiban's risk score in a head-to-head test on real number bans.
03
Prove agencies will pay for client-shareable reports, not just pacing.

[ Inspired by ]What it borrows, and what's different

Mitra MCP
Lets Claude and ChatGPT act through your own phone number, so recipients answer. Mitra is voice only, built for iPhone, and per-user. Canary covers the WhatsApp side Mitra doesn't, targets builders, and adds cross-user risk data.
DialMCP
Own-number agent calling with mandatory AI disclosure, transcripts and hard anti-abuse limits. DialMCP's limits are fixed rules set by the vendor. Canary's limits are learned from pooled ban outcomes, and Canary works on messaging in India rather than US/Canada calls.
WAHA (WhatsApp HTTP API)
Free, self-hosted REST API over WhatsApp. It is the default way builders put agents on their own number. WAHA ships the connection and leaves pacing and risk to the user. Canary wraps it as an out-of-band monitor and governor with shared outcome data, and doesn't replace it.

[ Sources ]11 cited

Mitra MCP launched on 10 Nov 2026 so agents can call from the user's own number, which confirms demand for agents acting through personal lines. It covers voice only.fourweekmba.com ↗
DialMCP launched in July 2026 for calls from the user's own verified number. It covers US and Canadian numbers only, which leaves India and messaging uncovered.prnewswire.com ↗
Meta's Business Solution Terms bar AI providers from offering a general-purpose assistant as the primary functionality from 15 Jan 2026. Official channels are closed to personal assistants.respond.io ↗
For a personal assistant on your own WhatsApp, the Business API cannot do it and the unofficial QR route is the only option. Ban risk is acknowledged as real.openclaw-easy.com ↗
Personal accounts running automation are described as an explicit detection target. Bans are reported as instant and irreversible, and the number can't be re-registered.pipes.bot ↗

Why it failed

Check 3 · Already exists

What already exists

6 products checked

ProductLive number risk meterHuman-like pacing governorAnonymous pooled benchmarksFailover and owner alertsShareable number-health reportOverlap with idea

This ideaWhat Canary plans to do

5 / 5

baileys-antiban (kobie3717)Free, drop-in TypeScript library for Baileys that adds warm-up, pacing, a risk meter, auto-pause and alerts. · the reason it failed

Unofficial WhatsApp agent tools can get your number banned, and nobody knows which behaviours cause it.

Canary sits between your agent and WhatsApp, showing your number's risk and pacing sends to reduce the chance of a ban.

Why it failed

What already exists

Unofficial WhatsApp agent tools can get your number banned, and nobody knows which behaviours cause it.

Canary sits between your agent and WhatsApp, showing your number's risk and pacing sends to reduce the chance of a ban.

Why it failed

What already exists

How it was checked