Agentic AI Comparison:
Project Mariner vs Raya by Teammates.ai

Project Mariner: 9

Project Mariner is explicitly described as a browser-native autonomous web agent that can plan and execute multi-step tasks in Chrome with minimal human intervention. It sees what is on the screen (text, images, forms, code), reasons about page state, and then performs actions—clicks, scrolls, navigation, form filling—based on high-level user goals such as "book a flight" or "find a table at 8 pm near Koramangala under ₹1,500". Mariner can carry out complex workflows end-to-end, including search, comparison, checkout flows, and bookings, while keeping logs and occasionally asking clarification questions when needed. However, it is constrained to browser-only control (no full desktop or system automation) and is framed as a research prototype that maintains user oversight via confirmations for sensitive actions. This combination of high autonomy within the browser but limited scope beyond it justifies a strong autonomy score of 9 rather than 10.

Raya by Teammates.ai: 9

Raya is marketed as a fully autonomous AI teammate that "manages your entire support operation" and "resolves tickets end-to-end" across multiple channels. It operates 24/7, responds across chat, email, phone, WhatsApp, Slack, Teams, Instagram, Facebook, and live chat, and takes real operational actions through 30+ native integrations, such as processing refunds, handling payments, updating CRM records, and changing account settings. Teammates.ai documentation and independent reviews emphasize that Raya is not a simple FAQ bot but a full AI agent that can resolve 78–90%+ of tickets without human intervention, escalating only when it is stuck or when policies require human oversight. The presence of enterprise controls, credit-based limits, and escalation pathways indicates strong but controlled autonomy within the customer-service domain rather than unconstrained general autonomy. Given Raya’s end-to-end autonomy in its domain but narrower task variety compared to a general web agent, a score of 9 appropriately reflects its high level of autonomous operation.

Project Mariner: 7

Project Mariner is designed to be invoked directly from Chrome, effectively acting as a screen-aware assistant: users describe goals in natural language and Mariner plans and executes actions on the visible pages. This interaction paradigm—"tell the agent what you want, watch it work"—is conceptually easy for non-technical users, and Mariner provides real-time logs, action histories, and prompts for confirmation on critical steps. However, Mariner is described as an experimental research prototype, initially limited to Google AI Ultra subscribers and select testers, with constraints such as U.S.-only availability and browser-specific setup (Chrome extension, AI Ultra plan subscription). Reviews and explanatory guides highlight that while Mariner is powerful, some workflows may require users to understand how to phrase multi-step goals, monitor logs, and manage permissions, which can raise the learning curve relative to mature SaaS products. Therefore, the user experience is promising but not as polished or broadly accessible as typical commercial support platforms, warranting a moderate-high ease-of-use score of 7.

Raya by Teammates.ai: 8

Raya is delivered as a managed SaaS product with guided configuration across channels (chat, email, WhatsApp, social) and out-of-the-box integrations into common support tooling. Teammates.ai provides a solutions directory, help center articles, and step-by-step configuration guides that explain how to connect channels, map workflows, and define policies, reducing the setup friction for typical business users (support managers, operations teams). Pricing pages emphasize that all plans, including Free, give full access to Raya with no feature gating; credit-based usage metrics (e.g., 1 credit per ~10 support replies) are spelled out to make understanding costs and limits straightforward. Independent reviews further note that Raya behaves like a drop-in replacement or augmentation for existing support teams, with analytics and ROI dashboards available on business tiers to track performance. The main complexity stems from deeper integration and configuration for multi-tool environments (Zendesk, Salesforce, 50+ tools), which requires some operational know-how but is typical of enterprise SaaS. As a result, Raya achieves high ease of use for its target business audience, scoring 8.

Project Mariner: 9

Project Mariner’s core design is general-purpose web interaction: it can interpret arbitrary web pages, reason over complex interfaces, and execute sequences of actions on almost any site within Chrome. It reads the DOM and accessibility tree, understands forms, text, images, and other elements, and can dynamically adapt workflows—e.g., navigating multiple comparison sites, updating search filters, filling multi-step checkout processes, or handling non-standard booking flows. Documentation and third-party analyses emphasize that Mariner is not limited to a single domain like e‑commerce or travel; instead, it acts as a general web agent for online shopping, research, information retrieval, bookings, and generic browser tasks. Its constraints come from its browser-only focus (no direct access to native apps or back-office tools outside the browser) and its status as a research prototype with guardrails and limited customizability towards specific enterprise workflows. Nonetheless, as a general web automation agent, its task flexibility is higher than most domain-specific agents, supporting a score of 9.

Raya by Teammates.ai: 8

Raya is highly flexible within the customer service domain, supporting omnichannel interactions (phone, email, chat, WhatsApp, social) and integrating with a wide ecosystem of CRM, ticketing, and business tools. It can handle a broad variety of support tasks: order tracking, refunds, returns, account changes, subscription management, payment handling, and policy-based responses, while operating in 50+ languages including many Arabic dialects. By design, Raya shares context with Adam (sales) and Sara (hiring), enabling cross-functional workflows (e.g., support plus upsell or support plus recruitment updates), which adds organizational flexibility. However, its capabilities are tightly focused on business support workflows and integrated back-office operations, not on arbitrary web browsing or unrestricted multi-domain tasks. Configuration guides and pricing also frame Raya as a tool specialized for customer service, scheduling, and related communication-heavy use cases. Consequently, Raya achieves strong flexibility within its specialization but remains less general-purpose than a browser-native agent like Mariner, resulting in a score of 8.

Project Mariner: 4

Project Mariner was made available as part of high‑end Google AI subscriptions such as the Google AI Ultra plan, with limited regional availability (notably U.S.-only at first). Documentation and guides note that access required subscription to premium AI tiers, with some sources citing monthly costs on the order of $249.99 for AI Ultra, positioning Mariner as a feature for a high-priced plan rather than a standalone affordable product. As a discontinued research prototype, Mariner does not currently offer a public, stable pricing structure; instead, its technology has been folded into Gemini Agent and other Google features, which may be priced at enterprise or consumer tiers unrelated to Mariner’s original experiment. From a cost perspective, this combination of premium access requirements, limited availability, and lack of affordable standalone tiers makes Mariner comparatively expensive or inaccessible for typical small-to-medium businesses. Given the high-cost and experimental framing, a score of 4 reflects its relatively poor cost profile compared to competitive SaaS offerings.

Raya by Teammates.ai: 9

Raya is offered on transparent, tiered pricing with a Free plan at $0/month that includes 10 credits (~100 support replies), and paid plans starting at $25/month (Pro) up to $100/month (Scale), plus custom Enterprise options. All plans include full access to all three teammates (Raya, Adam, Sara), omnichannel support (email, chat, SMS, voice, WhatsApp, social), 24/7 autonomous operation, and support for 50+ languages, with credits governing usage rather than per-seat or per-feature fees. Independent pricing analyses emphasize that 1 credit buys roughly 10 complete support replies for Raya, making cost estimation straightforward and generally cost-effective relative to hiring human agents for equivalent ticket volumes. The absence of feature gating—"every feature included"—and lack of hidden per-seat costs further enhances cost-effectiveness, especially for small businesses and growth-stage companies. Taken together, Raya’s accessible entry point, clear scaling path, and high automation ratio per dollar justify a high cost score of 9.

Project Mariner: 6

Project Mariner received substantial media coverage and industry attention when it was unveiled, with detailed articles from major outlets and AI-focused blogs highlighting it as one of Google DeepMind’s most ambitious consumer agent experiments. It was showcased at Google I/O and other events, contributing to high visibility among AI researchers, developers, and early adopters interested in agentic web browsing. However, Mariner was limited to small groups of testers and Google AI Ultra subscribers, remained a research prototype, and was ultimately discontinued as a standalone product in May 2026. Directories and comparison sites list Mariner as an experimental agent with limited user base and primarily research-oriented usage rather than mass-market adoption. This combination of high visibility but constrained access and short product lifespan suggests moderate popularity rather than widespread adoption, warranting a score of 6.

Raya by Teammates.ai: 8

Raya is positioned as Teammates.ai’s flagship product and is prominently featured across the company’s marketing, solutions directory, and pricing, as well as third-party reviews of AI customer-service platforms. It has been available since around 2024 and continues to be actively developed, with frequent references to its autonomous support capabilities, integration ecosystem, and performance metrics (e.g., resolving 78–90%+ of tickets autonomously). Reviews and comparison guides position Raya alongside and sometimes against major incumbents such as Zendesk AI, indicating growing traction in the customer-service space. Teammates.ai’s pricing and competitors pages also frame Raya as part of a broader competitive landscape, suggesting a notable existing customer base. While it may not yet match mainstream awareness of consumer chatbots or large enterprise platforms, Raya’s active commercial deployment, positive third-party coverage, and integration footprint indicate above-average popularity in its niche.