Project Mariner

Google's web-browsing agent inside Chrome

Agents & Automation Free
Researched · Published
RECATOOLS Score
4.2 / 10
Capability
6
Value for money
4
Ease of use
5
ASEAN readiness
3
API quality
3
Founded
2024
HQ
Mountain View, California, USA
Users
Launched
Developer

Overview

Project Mariner is Google DeepMind's web-browsing agent — operates a Chrome browser to complete multi-step tasks via Gemini. Distinguished by Google's native browser integration. Available to select testers; broader rollout planned.

Advertisement

Pricing

Pricing shown for reference only. These figures reflect RECATOOLS research as of 19 May 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.

Free
Free
Free tier with core features.

Use cases

Web agent Browser automation Multi-step tasks

What you can produce with Project Mariner

  • Delegate a multi-step web errand — such as researching a service provider and booking an appointment — and have the agent navigate real websites, fill forms, and pause to ask you for personal details before submitting (now via Gemini Agent, which inherited Mariner's technology).
  • Run browsing tasks on cloud virtual machines in the background so your own browser stays free while up to ten agent jobs execute simultaneously.
  • Teach the agent a repeatable web workflow once and replay it, a Mariner capability Google carried into its Gemini agent stack.
  • Let Chrome's Auto Browse feature, built from Mariner's algorithms, click through complex web flows like multi-page forms without manual input.
  • Ask AI Mode in Google Search to complete agentic tasks such as checking availability or comparing options across sites, another surface where Mariner's technology now lives.
  • Watch and interrupt the agent mid-task at any point, a human-oversight design Mariner pioneered and its successors retain.
Advertisement

ASEAN Perspective

Project Mariner in Southeast Asia

ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).

RECATOOLS Verdict

Project Mariner was Google DeepMind's research-preview AI agent that could navigate websites and complete multi-step web tasks (forms, research, shopping) much like a human, an important early demonstration of agentic browsing. As of May 2026 it has been discontinued as a standalone product, with its core technology folded into Gemini Agent and Chrome's Auto Browse feature.

Because it no longer exists as an independent tool, there is nothing to adopt directly: anyone interested should evaluate Gemini Agent and Chrome Auto Browse instead. Historically it was a capable, ambitious agent but always a gated experiment with limited access. ASEAN relevance is effectively nil now, the successor Auto Browse launched US-only to AI Pro/Ultra subscribers, so regional availability of the technology remains restricted and tied to Google's broader Gemini rollout.

Independent AI-assisted assessment by RECATOOLS.

What people say

Project Mariner no longer exists as a standalone product. Google officially shut it down on May 4, 2026, ending a roughly 17-month run that began with its December 2024 unveiling alongside Gemini 2.0 and a May 2025 rollout to paying subscribers. The technology was not abandoned: Google folded Mariner's browsing-agent capabilities into Gemini Agent (the task-automation layer of the Gemini assistant), into AI Mode in Search, and into an Auto Browse feature that arrived in Chrome in early 2026. Anyone looking for Mariner today should look at those surfaces instead.

While it was alive, tester feedback was consistently mixed. Reviewers who got hands-on access praised the concept — a Gemini-driven agent that could operate a real Chrome browser, fill forms, and complete multi-step tasks like finding and booking a plumber, pausing to ask for details such as name and email before proceeding. But the same reviewers found it slow at identifying where to click, prone to getting stuck in loops on tasks like completing a flight booking, and easily derailed by CAPTCHAs and login walls. TechCrunch and other outlets described the prototypes as slow and mistake-prone, and reviewers admitted to speeding up their demo videos to make the experience watchable.

Google did iterate on that feedback, moving Mariner onto cloud virtual machines so it could run up to ten tasks in the background while users worked on other things. But access remained gated behind the expensive AI Ultra subscription tier, and the standalone brand never reached general availability before being retired.

As a directory entry, Mariner is now primarily of historical interest: it marked Google's first serious public attempt at an autonomous web agent. Users who want that capability from Google today should evaluate Gemini Agent and Chrome's Auto Browse, which inherited its core algorithms.

Summary of public user & expert reviews, compiled by RECATOOLS.

About this listing

Researched on
Published on

This entry was compiled from publicly available data including Project Mariner's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Project Mariner unless explicitly stated.

Data accuracy

Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.

For the latest details, please refer to Project Mariner directly →

Spotted something out of date? Suggest an update →

Advertisement