The archive · Developer & Business Tools · Product decision · 2026
AnySearch bets agents need machine-first search: PH Weekly #1, 100k devs in a month
Chinese team's agent-only search infrastructure launched May 11, 2026; 100k developers in month one, 4M+ searches, then Product Hunt Weekly #1 on July 13, 2026.
AnySearch
What the business is
AnySearch is agent-only search infrastructure: it understands a task's intent, routes to 20+ vertical data sources (finance, legal, academic, security, code, corporate), and returns citation-backed structured Markdown an agent can use directly via Skill, MCP or API.
How it started
Four AI developers who had spent a year building with agents founded AnySearch after repeatedly hitting the same wall: a broad information request needs many searches, and professional data (financial filings, court records, academic citations, threat intel) sits outside ordinary search engines. They interviewed 100+ developers, distilled three needs — fewer calls, domain data, workflow-shaped results — and launched the product on May 11, 2026.
What happened
The open-source anysearch-skill hit #1 on Skills.sh within a week; in its first month AnySearch passed 100,000 developers, 4M+ searches and 4,000 GitHub stars across Asia-Pacific, North America and Europe. On July 13, 2026 it took Product Hunt's Weekly #1 — the first search product in a year to top a leaderboard dominated by agent and model launches — after releasing V2.1.0 on June 3 with rebuilt ranking and routing.
No ending yet — it is still running.
Background
AnySearch is search infrastructure built for AI agents rather than people. It takes a query, identifies the underlying task (e.g. company due diligence), routes to specialized sources across 20+ verticals — corporate registries, legal databases, academic citations, threat-intel feeds, code repositories — then returns citation-backed structured Markdown an agent can consume directly through a Skill, MCP or API.
The four-person founding team, all AI developers, built it after a year of watching agents struggle with search designed for human eyes: many round trips, token-heavy context, and professional data ordinary engines never index. They interviewed 100+ developers before shipping, launched on May 11, 2026, and the anysearch-skill repo hit #1 on Skills.sh within a week.
In its first month AnySearch passed 100,000 developers and 4,000 GitHub stars with 4M+ cumulative searches across Asia-Pacific, North America and Europe. On July 13, 2026 it ranked #1 on Product Hunt's Weekly Leaderboard — the first search-focused product in a year to break the AI-agent and foundation-model run — after only two months on the market.
Revenue plans are API and enterprise pricing plus a Pro tier; the product is free for individuals at 1,000 searches/day, with a zero-tracking, zero-telemetry architecture. The founders compare themselves to Exa (which raised $250M at $2.2B in May 2026) but bet on professional data beyond the indexable web as the wedge that makes agents pay.
What has to be true
- It bet on an audience incumbents ignore — agents, not humans — so a two-month-old product could win Product Hunt's weekly top without fighting search incumbents head-on.
- The Skill/MCP/API packaging rode existing agent ecosystems (GitHub, ClawHub, Skills.sh) to near-zero-friction adoption and organic distribution.
- Structured, source-attributed output attacks a measurable cost: in the team's test, AnySearch finished a code-research task in 1 search call versus 7, 16 and 28 for other tools.
- Self-building high-value data pipelines (finance, legal, security) is a moat answer: data quality and freshness stay in-house rather than behind a third-party API.
What can be applied
Targeting an audience nobody else serves — agents, not humans — let a two-month-old product win a global leaderboard; the test is whether devs pay for information quality once the free tier ends.
Aftermath
As of September 2026 AnySearch is live and still growing through developer channels: the product remains free at 1,000 searches/day for individuals, with Pro and Enterprise plans listed on its site and revenue planned from APIs and enterprise deployments. The team released V2.1.0 on June 3, 2026, its largest algorithm and architecture upgrade since launch, and says its next milestone is becoming the default information layer inside agent workflows. The open question it is testing: whether developers will pay for professional-domain information quality once the free tier ends.
Sources
- AnySearch Tops Product Hunt Weekly Leaderboard as AI Search Infrastructure Gains Momentum
- 瞄准 Agent 信息搜索层，中国团队AnySearch凭什么登顶Product Hunt？
- 上线一月吸引10万开发者，这家中国团队帮Agent看见网页外的世界
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card