Retriever AI Free Mode: What's Included vs. What Needs a Subscription
Retriever AI's free mode covers browser agent tasks with sponsored ads, but scheduling and remote MCP control still require the $9.99/mo starter plan.
What does Retriever AI’s free mode actually include?
Retriever AI’s free mode lets you run its Chrome extension and cloud browser agent without spending credits, as long as you’re willing to see a sponsored card alongside your results. The agent covers the AI processing and cloud browser time for eligible runs, subject to daily fair use limits. Anything beyond that allowance, such as enrichment services, still draws from a credit balance. In practice, this means a single person doing occasional browser automation work (scraping product pages, comparing listings, filling out forms) can do so without paying anything, provided the task fits inside the daily cap.
This matters because browser automation agents have mostly lived behind paywalls or free trials with hard expiration dates. Retriever’s approach is different: free access is ongoing, funded by ads rather than a time-limited trial, but it comes with real limits on volume and on which features are exposed.
TL;DR
- Free mode covers core agent runs, meaning AI reasoning and cloud browser time for eligible tasks don’t consume credits, though a sponsored card shows up with each result.
- Daily fair use limits apply, so free mode is built for individual, one-off tasks rather than continuous or high-volume scraping jobs.
- Skill files are part of the free experience, letting the agent save markdown notes about your preferences and reuse them on similar future tasks.
- Shared website skills are maintained across users, meaning improvements one person’s session produces (like finding a hidden filter menu) can benefit everyone using the same site.
- Scheduling, triggers, and remote MCP control of the browser extension require the paid starter plan, currently listed at $9.99 a month with a monthly credit allowance.
- Connecting outside MCP servers to give Retriever new tools is separate from letting another agent control Retriever, and only the second capability is gated behind a subscription.
One coffee. One working app.
You bring the idea. Remy manages the project.
How does the free browser agent actually work?
You give Retriever a task in plain language: what you want done, which sites to use, and what the finished output should look like. Typical jobs include building a comparison spreadsheet from several product pages, screening job listings against a resume and filling out applications, or working through a homework problem step by step. The agent opens pages, reads content, fills forms, and assembles the result according to your instructions.
To start, you install the Chrome extension, sign in, and switch on free mode. From there you can launch a task directly from whatever page you’re on, which is useful for sites you’re already logged into. There’s also a cloud dashboard version if you’d rather not run everything through the browser extension.
A practical example: comparing microphones across several open product pages. A prompt like “make a table with model, listed price, connection type, and product link, leave anything you can’t find blank” gives the agent clear boundaries and tells it what to do when data is missing. The output still needs a human check, especially around details like whether the agent grabbed a sale price instead of the regular price, or picked the correct product variant among similar listings.
What are skill files and why do they matter?
One of the more distinctive parts of Retriever is its skill file system. These are markdown files the agent writes and updates as it works, capturing durable lessons like formatting preferences or recurring workflow corrections. If you tell it every comparison table should include a source link and leave missing prices blank, that preference gets saved and applied automatically to your next similar task, without you repeating the instruction.
These personal skill files are readable and editable, and you can open a skills library to see what’s been saved. The agent is designed to only write a new skill when it learns something that will actually help future tasks, rather than logging notes after every trivial run.
Separately, Retriever also maintains shared website skills across its user base. These capture knowledge about how a specific site behaves, like a filter option buried inside a menu that the agent might otherwise miss. When one user’s session teaches the agent how to navigate that quirk correctly, the lesson can be preserved and made available for future runs on that site, including for other users. The team updates these as websites change. This is a meaningful design choice: personal skills describe how you want work done, shared skills describe how a given website actually works, and combining the two is meant to reduce repeated mistakes over time. The agent still has to apply the guidance correctly each time, so it’s not a guarantee of perfect execution.
What still requires the paid starter plan?
- ✕a coding agent
- ✕no-code
- ✕vibe coding
- ✕a faster Cursor
The one that tells the coding agents what to build.
A few capabilities sit outside free mode entirely. Background schedules and triggers, meaning tasks that run automatically on a timer or in response to an event (checking a website every morning, for instance) require a subscription. Manually kicking off a supported task stays free, but automating that same task on a recurring basis does not.
The other paid feature involves MCP (Model Context Protocol), which is a way for an agent to reach tools hosted by another service, such as a database or internal system. Connecting an outside MCP server so Retriever can use its tools is available in the extension settings under the tools section: you add the server’s URL, authenticate if needed, and its tools become available to the agent. That part isn’t gated.
What does require payment is the reverse setup: exposing the Retriever browser extension itself through MCP, so a separate AI agent can control your browser through it. That remote browser access is a subscription feature. It’s worth keeping these two directions distinct, since they sound similar but serve different purposes: one lets Retriever pull in outside tools, the other lets outside agents drive Retriever.
Is Retriever AI’s free mode worth it?
For individual, occasional tasks, the free tier is a reasonable way to test whether a browser agent fits into your workflow before paying for anything. Scraping a handful of product pages, prepping a batch of job applications, or working through study questions are all tasks you can try without a subscription, limited only by the daily fair use cap and the sponsored card that appears with results.
Where the free tier stops making sense is recurring or automated work. If you want a task to run every morning without you starting it, or you want another AI agent to operate the browser extension remotely, you’re into paid territory. The starter plan is listed at $9.99 a month and includes a monthly credit allowance for paid usage beyond the free allotment.
A sensible way to evaluate it: start with a task you already do by hand regularly, run it through free mode, and check the output carefully. If the agent handles it reliably and you find yourself wanting it to run unattended or on a schedule, that’s the point where a subscription starts to pay for itself.
Frequently Asked Questions
Does Retriever AI’s free mode cost anything?
No signup fee or subscription is required to use free mode, but a sponsored card appears alongside results, and daily fair use limits cap how much you can run before hitting a wall.
What happens if I exceed the free mode limits?
Extra services beyond the included daily allowance, such as enrichment features, draw from a credit balance, which requires being on a paid plan with credits available.
Can I schedule tasks to run automatically for free?
No. Background schedules and triggers that run tasks automatically on a timer or event are part of the paid starter plan, not free mode.
What’s the difference between connecting MCP tools and exposing Retriever via MCP?
Connecting an outside MCP server gives Retriever new tools to use during its own tasks, and that’s available in free mode. Exposing the Retriever browser extension through MCP so another agent can control it remotely is a separate, paid capability.
Are skill files available in free mode?
Plans first. Then code.
Remy writes the spec, manages the build, and ships the app.
Yes. Personal skill files and shared website skills are part of the core agent experience and aren’t listed as a paid-only feature.
