AI SDR

A multi-tenant AI-powered Sales Development Representative platform for B2B outbound prospecting, personalized outreach, and campaign execution. The system uses a CrewAI multi-agent workflow for campaign execution and supports email sequences, knowledge extraction from documents, and integrations with Gmail and Google Sheets.

Features

• CrewAI workflow: Four-agent crew (Prospector, Personalization, Outreach, Coordinator) runs on campaign execute. The crew validates leads, generates personalized copy, and coordinates the pipeline via sequential tasks.
• Campaign and lead management: Create campaigns, add or upload leads (CSV), and track status through the UI and API.
• Specialized agents: Prospector (lead generation), Enrichment (validation and scoring), Copywriter (message personalization), and Smart Campaign Orchestrator (prompt to campaign with quality gates).
• Email sequences: Multi-step sequences (email, delay, condition, action) with enrollment and scheduled execution. Steps run on an interval; email steps send via Gmail or SMTP.
• Knowledge base: Upload company or sales documents; the system extracts structured knowledge with Claude and uses it for campaign suggestions and copy context.
• Integrations: Google OAuth (Gmail, Sheets, Drive), SMTP/IMAP. Optional Grok for market intelligence.

Tech stack

• Backend: FastAPI, Supabase (PostgreSQL), APScheduler for sequence execution
• Frontend: Next.js 14, NextAuth (credentials and Google)
• AI: CrewAI, OpenAI, Anthropic (Claude)

Prerequisites

• Python 3.8+
• Node.js 16+
• A Supabase project (schema in supabase_schema.sql; run migrations as needed)

Setup

1. Clone the repository and create a virtual environment:

   python -m venv venv
   source venv/bin/activate   # Windows: venv\Scripts\activate
   pip install -r requirements.txt
   

2. Copy environment configuration and set variables:

   cp env.example .env
   

   Required for the Supabase backend: SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY. For AI features: OPENAI_API_KEY, ANTHROPIC_API_KEY. For Google (Gmail, Sheets): GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET, GOOGLE_REDIRECT_URI. See env.example for the full list.

3. Frontend:

   cd frontend
   npm install
   cp env.local.example .env.local
   

   Set NEXTAUTH_SECRET, GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET, and NEXTAUTH_URL (e.g. http://localhost:3000) in .env.local.

Running the application

Backend (Supabase, full API and sequence scheduler):

python backend/main_supabase.py

Serves at http://localhost:8000. API docs at http://localhost:8000/docs.

Frontend:

cd frontend && npm run dev

Serves at http://localhost:3000. Set NEXT_PUBLIC_BACKEND_URL to http://localhost:8000 if the backend runs elsewhere.

Sign in with the configured credentials or Google OAuth. Create a campaign, add leads, and use Execute to run the CrewAI workflow. Sequences, knowledge extraction, and other features are available from the dashboard and campaign pages.

Repository structure

• agents/ – CrewAI workflow (workflow.py), Google-integrated pipeline (google_workflow.py), and specialized agents (Prospector, Copywriter, Enrichment, etc.)
• backend/ – FastAPI app (main_supabase.py for Supabase; main.py for SQLite)
• frontend/ – Next.js app and API routes
• integrations/ – Email, Google OAuth, Google Sheets, LinkedIn (stubs), Grok
• services/ – Supabase, knowledge, sequence execution, campaign intelligence

For a detailed analysis of what is implemented and how the system works, see REPO_ANALYSIS.md.