Anvil is an AI-powered lead generation and e-commerce operations platform. It discovers high-intent leads across 20+ social platforms, automates outreach campaigns, and unifies your CRM, pipeline, and ERP in a single workspace.

Which platforms does Anvil monitor for leads?

Xiaohongshu, Douyin, Weibo, Bilibili, WeChat, Taobao, JD, Pinduoduo, Instagram, TikTok, Twitter, Facebook, YouTube, LinkedIn, Reddit, Pinterest, Amazon, Shopify, and more — 20+ in total.

Is there a free plan?

Yes. Free tier includes 500 lead collections, 1 AI bot, and access to the core CRM. Upgrade to Starter ($49/mo) for higher quotas, Pro ($199/mo) for AI agents and outreach automation, or Enterprise for custom limits.

Does Anvil support Chinese platforms like Xiaohongshu and WeChat?

Yes. Native integrations for Xiaohongshu, Douyin, Weibo, Bilibili, WeChat (work + personal), Taobao, JD, and Pinduoduo. Built specifically for Chinese B2B and cross-border e-commerce teams.

Which AI models does Anvil use?

Anvil uses DeepSeek v4 for chat, agents, reasoning, and content generation, with OpenAI text-embedding-3-small for vector search/RAG. ElevenLabs provides multilingual voice synthesis.

Docs/Features/Lead Scoring

Lead Scoring

A 0-100 score per lead computed from intent signals, fit signals, and engagement signals. Drives Sequence enrollment, AI Employee task priority, and your own routing rules.

![Lead Scoring](/docs-img/lead-scoring.png)

What It Does

Combines fit (ICP match) + intent (engagement) + recency into a single score.

Updates the score in real-time as new activities arrive.

Emits `lead.score_changed` webhooks when the score crosses a configured threshold.

Tiers leads (`HOT`, `WARM`, `NURTURE`, `COLD`) for one-glance triage.

When to Use

Always — the score is the single most useful column in your CRM. Sort by score descending to focus your day.

Score Components

Component	Weight	Example signals
Fit	30%	Industry match, employee count, country
Intent	50%	Replies, opens, web visits, pricing-page hits
Recency	20%	Days since last interaction (decays exponentially)

Each component is itself the output of a sub-model — you can override weights per-tenant in **Settings > AI > Scoring**.

Tier Mapping

Score	Tier
80-100	HOT
60-79	WARM
30-59	NURTURE
0-29	COLD

API Endpoints

`GET /api/v1/leads/:id` — score + tier in the response.

`POST /api/v1/leads/:id/recompute-score` — force a recompute (rare; the engine recomputes on every activity).

`GET /api/v1/leads?minScore=80` — filter by score.

Pricing & Limits

Plan	Real-time scoring	Custom weights
Free	Daily batch	No
Starter	Hourly	No
Pro	Real-time	Yes
Enterprise	Real-time	Yes

Gotchas

Cold-start: brand new leads with no activity score around 35-45 (fit-only). Don't read too much into low early scores.

The recency decay is tenant-configurable; default half-life is 14 days.

Score changes have a 30-second debounce — don't expect a webhook fire for every micro-event.

← Back to Docs