The POC Agent monitors every active proof-of-concept pilot in your pipeline by pulling deal data from HubSpot and querying product usage metrics from Metabase. It analyzes each company's engagement against your POC success criteria and posts a structured health summary to Slack — surfacing which pilots are on track, which are at risk, and what signals to act on.

Instead of manually cross-referencing CRM data with usage dashboards before each pipeline review, your GTM team gets a synthesized per-company brief delivered automatically to the channel where they already work.

Paste one prompt into Warp and the agent deploys itself:

Set up poc-bot: clone https://github.com/warpdotdev/poc-agent-oss.git if it
isn't already cloned, then run the setup skill.

The setup skill walks through the whole deployment: creating .env, collecting credentials, discovering your HubSpot and Metabase IDs via their APIs, and running read-only verification checks against each service. It is safe to re-run to finish a partial setup.

Or set it up manually:

1. Configure

Clone the repo, copy .env.example to .env, and fill in your values: HUBSPOT_ACCESS_TOKEN, the POC pipeline and stage IDs, your Metabase URL, API key, and dashboard IDs, and a Slack bot token (POC_BOT_SLACK_TOKEN) plus target channel. The README documents where to find every ID.

2. Load the environment

The scripts read from the process environment — there is no dotenv auto-loader — so export the variables before running anything:

set -a && source .env && set +a

3. Run the pipeline

# Fetch HubSpot deals + Metabase metrics
python3 .agents/skills/poc-analysis/scripts/fetch_poc_data.py

# Analyze, then post per-company health summaries to Slack
python3 .agents/skills/poc-analysis/scripts/post_to_slack.py

Once live, the agent runs on a scheduled cadence in Oz. Each run fetches fresh CRM and usage data, runs the analysis, and posts the latest health snapshot. Trigger manual runs for ad-hoc reviews or align the schedule to your pipeline review cadence.

Sales and solutions engineering teams running concurrent enterprise pilots benefit most — the agent eliminates the weekly pre-meeting data gathering that typically falls on SEs or sales ops. Every summary includes usage trends, milestone completion, and a risk signal so your team can prioritize follow-up and take targeted action before a pilot goes cold.

The agent is especially effective when POC success metrics are quantitative — active user counts, feature adoption rates, session frequency — because it can surface degradation trends days before they would otherwise appear in a manual review.