HubSpot releases Agent CLI in public beta for AI agents working on CRM data
HubSpot released the HubSpot Agent CLI in public beta on June 23, 2026. The company describes it as a command-line tool built for AI agents to read and change CRM data, suited to scheduled and bulk work.
By Cora Whitley, GTM AI Analyst, Chief of StaffHubSpot's June 23, 2026 changelog post introduced the Agent CLI, a binary separate from the HubSpot developer CLI. It is aimed at background work such as scheduled automations and bulk operations that run without a person in the loop. HubSpot positions its remote MCP server and AI connectors for conversational work and the Agent CLI for repetitive tasks. Output defaults to JSONL. Super admins can use App Governance, in beta, to control which users may connect the CLI to an account.
The details
- Status: public beta as of June 23, 2026.
- Named environments: Claude Cowork, Claude Code and OpenAI Codex.
- Commands follow a hubspot <noun> <verb> pattern across objects, pipelines, properties, workflows, activities and reports.
- Write commands support --dry-run.
- Admin mode uses a service key and does not create a per-user audit trail, according to HubSpot.
- Claude Team or Enterprise admins must allowlist api.hubapi.com for the CLI to run from Cowork.
Context
GTM Tech News covered HubSpot's launch of Agent Hub and Agent Builder in public beta on July 23, 2026, a month after the Agent CLI. The CLI adds a third AI access route alongside the Claude and ChatGPT connectors and the remote MCP server, which became generally available in April 2026.
Not disclosed
- A general availability date
- Rate limits for CLI operations
- Whether CLI usage counts toward HubSpot AI credits
- Vendor
HubSpot- Category
- CRM & Marketing Automation
- Announcement type
- AI launch
Report an error
Vendors and readers can flag a factual error in this story, or ask for an image or logo to be removed. Email Cora, the editor, with the link. Image and logo removals happen within one business day. For factual corrections, Cora replies with next steps and a ForgeX lead approves the fix before it is published.
