Not a chatbot. A research engine.
Sourcery is an autonomous research pipeline. Given any topic, it:
Builds a structured research plan → generates targeted multi-angle search queries → crawls and cleans real web pages (including anti-bot protected ones via residential proxies) → extracts structured claims → recursively follows citations and related sources → detects contradictions across sources, and → synthesizes everything into a fully citation-linked Markdown report — all without human input.
Every claim in the output is traceable to a real source fetched during that exact run. Citations aren't decorative — they're deterministically rebuilt from the actual source list. Sourcery doesn't smooth over conflicts; contradictions between sources are surfaced explicitly and carried into the final report.
How it compares to commercial research tools
Sourcery generates reports that are significantly more detailed and insightful than commercial deep research offerings from ChatGPT, Gemini, or Claude, while keeping every claim and insight fully grounded in evidence. Because it isn't tied to any subscription plan, it has no usage limits or paywalls. At under $0.10 per report — including the cost of residential proxies to bypass anti-bot measures — it offers unmatched value.
More detailed and insightful than anything produced by ChatGPT, Gemini, or Claude's deep research features.
Not tied to any subscription. No usage caps, no rate limits, no waiting for your quota to reset.
Under $0.10 — including residential proxy costs for anti-bot bypass. Market-leading, by a wide margin.
An autonomous research pipeline
- ✦Open source — code dropping on GitHub in the next few days
- ✦LLM-agnostic — works with any OpenAI-compatible inference endpoint
- ✦Web search via a third-party search provider
- ✦HTML extraction handled locally — no scraping SaaS
- ✦Residential proxy provider for anti-bot bypass
- ✦PDF parsing handled locally
- ✦SQLite persistence — all state (runs, sources, claims, contradictions, memory) stored locally
- ✦Resumable runs with cross-run source reuse