Browser-automation agent Tabstack purges page data after each call, discloses no compliance certification

Tabstack, a hosted browser-automation service reached through an API or an MCP command-line interface, describes itself as a way for an agent or application to hand off a plain-language task and have it carried out on a live web page. The vendor was promoted into AI Dispatch''s published index this week; its own site is the entity''s only source document to date.

"Describe a task in plain language and Tabstack carries it out on the live page: navigating, clicking, filling forms, and completing multi-step flows on sites you don''t control," the company''s site states. The model, browser and orchestration all run on Tabstack''s infrastructure — "The model, the browser, and the orchestration all run on Tabstack" — with no self-hosted, VPC or on-premise option described.

On data handling, the vendor commits to deleting what it collects: "Requests and retrieved pages are used to complete your call and support you, then purged. Never sold, never used to train models." On governance, the site states "You set the scope and depth of every call," putting the limit on each task in the caller''s hands; no spend ceiling or separate authority limit is described. An interactive mode "pauses for human input" when configured.

Pricing is "cost per action," billed in credits, with custom quotas, dedicated support and service-level agreements offered on an Enterprise plan — though no uptime target, measurement basis or remedy for missing one is published. The company''s only stated protocol integration is MCP, offered through an MCP CLI.

AI Dispatch''s index found no compliance certification, audit trail, insurance coverage or indemnification term disclosed anywhere on the page checked. For a service designed to act autonomously on sites the customer does not control, that leaves governance resting on the per-call scope setting alone, with no published backstop if something goes wrong.