Findr
Work search across Slack, Drive, Notion, and Jira — surfaced inside Slack and Chrome, where you already are. Co-founded, raised, shipped, wound down.
$70K revenue 1,000+ DAUs #1 Product Hunt 2023 → 2024
What it was
Findr connected to Slack, Google Drive, Notion, and Jira via OAuth, indexed the content, and answered work questions with sourced, streaming responses — inside the surfaces people already used. A Chrome extension made it the answer layer on top of any tab. A Slack app made it the answer layer inside the room you were already in.
Permission boundaries were preserved per source: you only saw what you already had access to, no more.
Why it existed
Your answer at work is in Slack, or Drive, or Notion, or Jira. You don't know which. So you ask three colleagues instead. Findr was the bet that one search across all of them — surfaced inside the tools people already use — was the obvious next step.
The call we got right
Distribution-first, not retrieval-first. The Slack and Chrome native clients meant Findr lived inside the surfaces people already used — no new app, no new tab, no context switch. The retrieval was good. The channel was the moat. We won the channels.
Where it broke
By 2024, the platforms had shipped overlapping primitives at platform scale. Slack AI, Notion AI, and Glean covered the same surfaces — with the platform's own data, permissions, and pricing power. The retrieval moat melted. We had won the channels and lost the platform shift.
We wound the company down rather than burn the seed on a fight we could not win.
What it shipped
- A unified search box that fanned out across Slack, Drive, Notion, and Jira
- Custom GPTs from your own documents — answer in your team's voice, on your team's data
- OAuth-per-source connectors with per-tenant scopes and refresh handling
- Streaming RAG — answers materialized token-by-token from across sources
- Native clients in the surfaces people already used: Slack and Chrome
- Per-source permission propagation — you only saw what you already had access to
- Self-host migration so enterprises could run Findr on their own infrastructure
- SOC2, ISO, GDPR, and HIPAA aligned by the time we shipped to enterprise
What it taught
- Distribution is the moat, not retrieval. Native clients carried the product further than search quality did.
- If your AI feature can be rebuilt by a platform owner over a weekend, you are renting their shelf space until they take it back.
- Horizontal AI features get absorbed into the platform layer. Win a workflow before the platform notices, or build for a vertical the platform will not.