Sage — The Alexa+ Sales Intelligence Layer

Your CRM that listens.

What is Sage?

Sage is a passive sales intelligence layer that sits on top of your existing CRM. It listens to your sales calls, extracts insights using Amazon Nova, proactively surfaces what you're forgetting, and syncs to your existing CRM via MCP.

Features

• Passive Listening — No wake words, no commands. Just have the conversation.
• Auto-Extraction — Contacts, deals, follow-ups, sentiment, buying signals extracted automatically.
• Proactive Insights — "You haven't followed up with Acme in 20 days."
• CRM Sync — One-way sync (Sage → CRM) with Salesforce, HubSpot, Pipedrive via MCP.
• Visual Pipeline — Deal boards, stuck-deal alerts, sentiment trends on Echo Show.

Quick Start

# Clone the repo
git clone https://github.com/DawnofGenX/sage.git
cd sage

# Start with Docker
docker compose up

# Or run locally
cd mcp-server
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
.venv/bin/pip install -e .          # required: installs src/ as a package
.venv/bin/python -m src.data.seed
.venv/bin/python -m uvicorn src.api.rest:app --port 8000
cd web-simulator && npm install && npm run dev

The Docker path needs no manual seed — entrypoint.sh seeds on first boot.

Running the tests

cd mcp-server
.venv/bin/python -m pytest tests/ -q     # expect: 298 passed

The pip install -e . step is not optional. The suite imports from tools.crud import ..., so src/ must be installed as a package; with dependencies alone you get collection errors (No module named 'tools') and zero tests run — see the verified 18-error negative case in docs/test-report.md.

MCP Tools

Tool	Description
extract_from_call	Process call transcript → structured CRM data
get_contact_context	Full contact history, deals, interactions
get_pipeline_health	Deal stages, stuck deals, sentiment trends
create_contact	Add new contact
update_contact	Update existing contact
create_deal	Add new deal
update_deal_stage	Move deal to new stage
schedule_followup	Create follow-up reminder
draft_followup_email	Generate follow-up email draft
log_call	Log a call record
get_deal_insights	Sentiment, risks, buying signals
get_daily_briefing	Morning summary
get_todays_followups	Prioritized action items
get_weekly_review	Weekly performance summary
search_contacts	Fuzzy search
sync_to_crm	Sync to external CRM
get_company_context	Company-wide contacts, deals, health
get_activities	Activity stream for a contact or deal
get_deal_history	Stage-change history for a deal
get_deal_timeline_events	Merged stage-change + activity stream, oldest first
get_deals	Every deal, ready for a pipeline board (with measured stuck status)
create_task	Create a task
enrich_contact	Enrich a contact with context
get_forecast	Weekly/monthly deal forecast
run_agentic_loop	Transcript → 5-step chained tool run over SSE

Architecture

See docs/architecture.md for the full architecture diagram.

Contributing

See CONTRIBUTING.md.

License

MIT

More writing on this stack: https://dev.to/dawnofgenx