Repo of the Day
Octosift/OctoSift
Published: Nov 19, 2026
Open repository ↗Reddit monitoring and lead generation tool, open source and self-hosted. Watches the subreddits where your customers ask for help, scores posts with your own AI key, and drafts replies and DMs you send yourself
Summary
OctoSift is a TypeScript, self-hosted Reddit monitoring and lead-generation app for people whose potential users ask for help in specific subreddits. It filters and scores posts, explains each verdict, and prepares reply or DM drafts for human review using an AI provider and key configured by the operator; it does not post for you.
What it is useful for
A developer can use OctoSift to find questions that match a product without manually reading every new thread. For example, the test suite covers an invoicing tool for freelancers, a consumer plant-care app, and a developer command-line tool. After OctoSift builds a product profile from a website, you confirm three to ten relevant subreddits. Posts scoring 60 or higher enter the inbox, while posts scoring 75 or higher also receive a draft.
The workflow is designed for review rather than blind automation. Each scan explains why a post was kept or rejected, and marking leads as good or bad turns those judgments into rules for later scans. Drafts are checked for whether they answer the question, disclose that you built the product, and avoid invented details. Subreddits marked as non-promotional receive drafts without product mentions. A private-message draft is also prepared when the author invites messages.
OctoSift does not watch all of Reddit; it reads only the selected subreddits. Requests are deliberately slow and can pause when Reddit pushes back. A connected Reddit account is recommended because public RSS is retired and remains a best-effort fallback. Scanning therefore takes time, and the README advises using a separate account for automated reading.
How engineers can use it
Start with the documented demo, which uses synthetic posts and a simulated provider and requires no key or network:
pnpm install
pnpm demo
Open http://127.0.0.1:3005. For a real deployment, the README documents:
cp .env.example .env
pnpm build
pnpm start
Then add an OpenAI, Anthropic, or OpenAI-compatible provider key, select a preset, connect Reddit, provide the product website, confirm subreddits, and run a scan. The dashboard also supports local providers such as Ollama and LM Studio. Budget caps stop scans at configured spending limits.
Engineers can inspect individual scans from the CLI with pnpm scan --once, including --dry-run and --json output. Development checks are available through pnpm check and pnpm test:e2e. Node.js 22.14+, 24, or 26 and pnpm are required. Docker users can run docker compose up --build; set APP_PASSWORD before exposing the dashboard beyond the local machine.