// the find
kaymen99/sales-outreach-automation-langgraph
Automate lead research, qualification, and outreach with AI agents and Langgraph, creating personalized messaging and connecting with your CRMs (HubSpot, Airtable, Google Sheets)
A LangGraph pipeline that pulls leads from HubSpot, Airtable, or Google Sheets, researches each company through its site, blog, social accounts, news, and a LinkedIn profile fetched via RapidAPI, scores them against criteria, and writes research, outreach, email, and interview-prep documents for the leads that pass. It is aimed at a small agency or freelancer who wants to adapt the pipeline to their own offer, and it ships with a sample agency and case studies to replace.
The workflow is split into explicit stages (fetch, research, qualify, outreach, CRM update), and qualification gates the outreach generation, so disqualified leads do not cost you the report-writing calls. The CRM integrations sit behind a base class in src/tools/leads_loader, so adding a CRM means writing one subclass rather than changing the graph. Outreach reports pull case studies through RAG over data/case_studies, and the sample outputs in reports/ show what a finished email and interview script look like before you run anything.
The last push was January 2025, and the README still describes Gemini models and a Dev.to walkthrough, so check that the pinned LangChain and LangGraph versions still install and run before you build on it. LinkedIn data comes from a third-party RapidAPI wrapper, which raises a terms-of-service question and adds a dependency that can change price or endpoint without notice. The Chroma index is committed as binary files under database/, which ties the vector store to the sample case studies and makes it awkward to version. Google Docs, Gmail, and Sheets all need OAuth credentials configured before the reports and emails reach anywhere, and that setup is a real chunk of work for a first run.