CLI wrapper that sends structured operational instructions to OpenAI Responses API with a Zapier MCP server tool attached.
You provide:
--org(organization context)--project(project context)--instruction(what to do)
The tool builds a strict prompt and lets the model use Zapier MCP tools when needed.
From inside hubspot/:
npm installOPENAI_API_KEYZAPIER_MCP_URL
Optional:
OPENAI_MODEL(default fallback isgpt-4.1)
export OPENAI_API_KEY="..."
export ZAPIER_MCP_URL="https://your-mcp-endpoint"
export OPENAI_MODEL="gpt-4.1"OPENAI_API_KEY=...
ZAPIER_MCP_URL=https://your-mcp-endpoint
OPENAI_MODEL=gpt-4.1- OpenAI key: from your OpenAI account API keys page.
- Zapier MCP URL: from your Zapier MCP/server setup or team admin.
If either value is missing, the script exits with a clear error.
node hubspot.js \
--org "Acme Inc" \
--project "Q1 Pipeline Cleanup" \
--instruction "Create follow-up tasks for stale deals older than 30 days"Helpful options:
--model <name>--tool-choice auto|required--approval never|...--json(prints full response JSON)--help
- This tool is best when your skill can gather org/project context first.
- Keep instructions concrete and auditable.
- Use least-privilege credentials for MCP-connected systems.