GTM Agents

Home for Warp's GTM (go-to-market) agents — purpose-built systems that help reps move from raw signals to scored, actionable outreach.

How to deploy

The fastest path: open this repo in Warp and paste this prompt into the agent input:

Run the setup skill

The agent picks up .warp/skills/setup/SKILL.md, which installs dependencies, validates the install with the test suites, walks you through .env configuration for the agents you care about, and points you at per-agent deployment docs.

Or follow the manual steps below.

Manual setup (reference)

Prerequisites

• Python 3.10+ (the fastmcp dependency requires it; CI runs 3.12).
• For live runs only: a Google Cloud project with BigQuery, a HubSpot private app token, and (for Slack digests) a Slack bot token. The test suites run with no credentials.

Install

python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

Configure environment

cp .env.example .env

.env is git-ignored. Every variable ships with a placeholder default (e.g. example-gcp-project) so the code imports and tests cleanly with no configuration; set real values to run against your own infrastructure. For Google Cloud auth, either set GCP_SERVICE_ACCOUNT_JSON or use Application Default Credentials (gcloud auth application-default login). See .env.example and CONTRIBUTING.md for the full variable reference.

Run tests

# Shared hubspot_agent + PLG tests
python3 -m unittest discover -s tests -t .

# BDR agent tests (src layout)
PYTHONPATH="$PWD/bdr_agent/src" python3 -m unittest discover -s bdr_agent/tests -p 'test_*.py'

Per-agent setup, configuration, and deployment instructions (warehouse mapping, HubSpot/Slack one-time setup, Oz cloud scheduling) live in each agent's README (plg_upsell/README.md, bdr_agent/README.md).

Agents

Agent	Status	Description
plg_upsell/	Live	Identifies PLG domains likely to convert to enterprise, surfaces top champion contacts, syncs scores to HubSpot, and posts a weekly digest to #your-plg-alerts-channel.
bdr_agent/	In progress	V1 outbound personalization. Researches HubSpot leads/accounts, generates source-backed personalized hooks, and writes them back to HubSpot for rep review. Phase 1 (HubSpot ingest) scaffolded.

Each agent owns its own scripts, SQL/dbt models, and reference material under its package directory. See the per-agent README for setup, configuration, and run instructions.

Shared

• hubspot_agent/ — shared HubSpot CRM library used by every agent in this repo (workflows, lists, properties, duplicate detection, etc.).
• tests/ — tests for the shared hubspot_agent library.
• .warp/skills/ — Oz cloud agent skills. Each skill is namespaced by agent.
• requirements.txt — Python dependencies shared across agents.
• .env.example — template for all supported environment variables; copy to .env and fill in.

Repo conventions

• Run commands and file paths in agent READMEs are repo-root-relative (e.g. plg_upsell/scripts/scoring_sync.py), so commands work from a single cd into the repo.
• Branch naming: <your-name>/<short-description> (e.g. jordan/feat-outbound-hooks-skeleton).
• New agents get their own top-level package (agent_name/) and a per-agent skill under .warp/skills/.