Answers for teams evaluating Aera

ChatGPT executes now too: agent mode drives real websites, and ChatGPT runs recurring scheduled tasks. It does that work on OpenAI's own virtual computer, so the browser is not yours and you sign in to sites by handing control to it. Aera does the same class of work in the Chromium profile you already use, in the sessions you are already signed into. The scheduler is a separate service on your machine, so a task still fires once the browser is closed, and every run keeps a step-by-step log you can read.

No wrapper. No overlay. No Chrome extension pretending to be more. Aera is a full browser with automation and MCP support built in. It can act directly on real web pages, and power users can inspect what's happening (like network traffic and logs) when they need to debug or build advanced workflows.

Reading, watching, and repeating are what it is dependable at. Pull the morning numbers out of a dashboard, sweep a set of sources, walk your own signup flow, triage an inbox, enter records into a portal, then do the same thing again tomorrow on a schedule. Ordinary form fields, search boxes, and text areas are solid. Rich text and code editors are not: in our own testing the agent damaged content inside them, so point it at a form rather than at a document editor. And one scoped job finishes far more often than a chain across ten unfamiliar sites, which is why we tell people to start with a single routine.

Those tools need code you write and then maintain. Aera takes a sentence. No scripts to debug, no selectors to update when a site ships a redesign, nothing to host. It re-reads the page on every run rather than replaying a stored selector, so a renamed div is not automatically fatal. And it runs in a browser you also use, not in a headless container somewhere else. Puppeteer still wins for deterministic test suites in CI.

You could. But that's another project, another codebase, another thing to maintain. Every time a website updates their UI, your automation breaks. Aera gives you the automation layer without the engineering overhead. Describe what you want. It runs. Your time is better spent shipping product, not maintaining browser scripts.

Minimal. If you can describe the job in plain English, you can use Aera. There is no scripting language and no configuration to fill in: you say what you need done, and Aera works out the how. The first routine is one sentence and a confirmation of the schedule. Start with a job you already do every morning, because you will know immediately whether the result is right.

Partly, and the split is worth knowing before you install. Aera can watch mentions, read threads, pull engagement numbers, track competitors, and draft replies on a schedule while you work on something else. Posting and sending stay yours: it prepares the reply, you press send. So it covers the research and triage half of a social hire, not the publishing half.

You do not have to sit and watch it, and it is not deterministic either. We would rather say both. The agent re-reads the page on every run instead of replaying a stored selector, so a moved button is usually survivable in a way a Puppeteer script is not. When a run does fail it stops, keeps the step-by-step log, and raises a notification. You read the log instead of a stack trace. What we will not tell you is that it never fails.

Time is your scarcest resource when you're bootstrapped. Every hour spent on browser busywork is an hour not spent on product, sales, or customer conversations. If Aera saves you even 1 hour daily, that's 30+ hours monthly. We think that's worth paying for.

It runs the watching half well and leaves the publishing half to you. Aera sweeps mentions across your channels on a cadence, reads the threads, filters out the noise, drafts replies, and hands you a digest with the drafts attached. What it does not do is post or send for you unattended. Rich composers are the surface where the agent is least reliable, and unattended mass replies are also the fastest way to lose an account. So the sweep is scheduled and the send stays a click you make.

Describe your ideal lead criteria. Aera can visit prospect profiles, analyze their activity, check company details, and compile qualified leads into a report, automatically. What used to take hours of manual research can run in the background while you're in meetings.

Absolutely. Schedule Aera to pull data from Google Analytics, social media dashboards, CRM, and any other web-based tool. It aggregates everything into a single report, formatted how you want it, delivered when you want it. No more morning data-entry rituals across 12 different tabs.

If your tools have a web interface, Aera can interact with them. It's not limited by API availability, Aera works through the browser, just like you would. Plus, MCP integration means you can connect Aera to your development workflow and build custom automations that span multiple tools.

Point it at competitor sites, social accounts, job boards, product pages, anything public. Aera checks on the cadence you set, compares against the last run, and surfaces only what changed, already summarized. You stop checking the same ten sources every morning and read a diff instead.

Your research, automations, and workflows can be organized by context. Switch between clients instantly without losing your place. The AI understands project context, so your research for Client A doesn't bleed into your work for Client B.

Time tracking says 8 hours worked. Actual productive output: maybe 5. The rest? Lost to browser overhead, context switching, and hunting for that tab you definitely had open somewhere. Aera gives you back those lost hours. More billable output in less calendar time, or more projects with the same capacity.

Aera is built for developers who need speed. Background automations run in separate processes and don't impact your active browsing. When you're debugging a frontend or running performance tests, Aera stays out of your way.

Your data stays on your machine. The inference does not, unless you run a local model. Both halves matter for client work, so here they are separately. Browsing history, bookmarks, and local data stay on your disk, and we have no access to them. An Aera-hosted model request sends what the agent reads (page content, your instructions, and the conversation) to a third-party AI provider, under their privacy policy. So on a page holding client data, either point Aera at a local model or accept that the page content leaves your machine. Our Privacy Policy and Terms of Service have the full detail.

Aera speaks MCP (Model Context Protocol). Connect it directly to Cursor, Gemini CLI, or any MCP-compatible tool. Your AI coding assistant can trigger browser automations, and Aera can feed data back to your development environment.

MCP (Model Context Protocol) is an open standard for connecting AI tools. It means Aera can talk directly to Cursor, Gemini CLI, Claude Desktop, and other developer tools. Your AI coding assistant can control the browser. Aera can feed real-time web data into your development workflow. No more being the middleware between your tools.

Yes. Power users can create reusable automation recipes and share them with the community. MCP support means you can also build custom integrations using tools you already know. No proprietary API to learn, MCP is an industry standard.

Aera supports local models through OpenAI-compatible endpoints such as Ollama, LM Studio, vLLM, llama.cpp, and Jan. Paid plans add Aera-hosted frontier models through OpenRouter, including GPT, Claude, Gemini, Grok, and Llama families, so you can optimize cost vs. quality per task.

The agent can read the pages and tabs you use it on so it can execute the workflow. For developers, Aera also supports deeper inspection (like network traffic) for debugging and building automations. You stay in control of what the agent runs and what data you keep.

Not today. Aera sits on open standards: Chromium underneath, MCP for tooling, and OpenAI-compatible endpoints for local models, so the routines you build are not locked to us. Opening parts of the automation layer is on the table and is not scheduled. If that decides it for you, say so at [email protected].

Headless automation (BrowserBase, etc.) runs in the cloud and is designed for backend processes. Aera is a full browser you use daily, with automation built in. Use it for browsing AND automation. No context switching between your 'real' browser and your automation environment.

Your data stays on your machine. The inference does not, unless you run a local model. That is the claim, and each half is checkable. We do not collect your browsing data, we do not train on your prompts, and we do not sell access to your behavior. History, bookmarks, and local data stay on your disk. Our business model is subscriptions, not surveillance, which makes you the customer rather than the product. When you use an Aera-hosted model, the page content, your instructions, and the conversation go to a third-party AI provider under their privacy policy, and we have no control over their handling once it leaves your device. A local model request goes only to the endpoint you configure. Our Privacy Policy and Terms of Service have the full detail.

On your machine. Browsing history, bookmarks, saved passwords, chat history, and task configs sit in local storage on your disk. We run no server that receives your browsing activity, so we cannot see what you visit, what you search, or what you automate. The model call is the exception. An Aera-hosted request sends your instructions, page content, screenshots, and conversation history to a third-party AI provider for processing, under their privacy policy rather than ours. A local model request goes only to the endpoint you configure. Our Privacy Policy and Terms of Service have the full detail.

That depends on which model you point at the page, and it is a choice you make per task. Your browsing data stays on your machine and we send no record of what you visit to any server of ours, which is why people handling proprietary code, client files, or competitive research use Aera at all. But an Aera-hosted request sends what is visible on the page to a third-party AI provider, and that includes the sensitive parts. For a genuinely sensitive page, run a local model. Then the page and your prompt go to your own endpoint instead of a provider. Aera still talks to our server about your account, a session check at launch, and updates; the security page lists exactly what those carry. Our Privacy Policy and Terms of Service have the full detail.

Aera operates like a normal browser, because it is one. We don't use techniques that trigger bot detection. The AI interacts with pages the same way you would, just faster. That said, respect each site's terms of service. Aera is a tool, how you use it is your responsibility.

We hold no compliance certification, so treat this as input to your own review rather than an answer. Browsing data stays on your machine, which helps. A local model request goes only to the endpoint you configure, so regulated content need never leave the device. An Aera-hosted request goes to a third-party AI provider under their privacy policy and their compliance posture, not ours. Read our Privacy Policy and Terms of Service, then take it to your legal and compliance team before you point Aera at regulated data.

Windows, macOS, and Linux, with full feature parity across all platforms. No 'Mac-first, others later' approach. Whichever OS you prefer, Aera works the same.

Yes. Aera supports Chrome extensions with direct import. Keep your password manager, ad blocker, dev tools, and everything else you rely on. No ecosystem rebuild needed, your existing workflow stays intact.

Aera runs on any modern system that can run Chrome. 4GB RAM minimum (8GB recommended for heavy automation use). No special hardware required, if you can run a browser, you can run Aera.

Install Aera, open the sidebar, and describe one job you already do by hand. Watch it run, then say "run this every morning" to make it recurring. Start with something you can check at a glance: a number from a dashboard, a page you re-read, a list you copy across. The skills marketplace has ready-made routines you can install and edit. If you get stuck, email [email protected].

The free plan is free for good, and it runs on a model you host yourself: you install a local server such as Ollama and point Aera at it in Settings > Models. There is no Aera-hosted model on the free plan, so if you would rather not run one, Pro is $20 a month and Ultra is $200 a month, and on those we host the models for you. Every plan needs an Aera account.

Fair question, and the answer you can check beats the answer we assert. Your data is a SQLite file on your own disk, not a row in our database. The browser is Chromium, so it keeps browsing whatever happens to us. Point it at a local model and no server of ours sits between the agent and the model. That part you can test in ten minutes. The rest is a promise, and ours is worth no more than anyone else's.

Email [email protected]. A person reads every message, usually the founder. Tell us what you asked Aera to do and what it did instead, and paste the run log from the sidebar if a scheduled task was involved. That log is normally enough for us to find the cause. There is a Discord for workflow ideas, but email is the channel we answer.

New versions ship every few days. Bug fixes deploy same-day. We're in rapid iteration mode, and your feature requests actually get built. This isn't enterprise software with annual updates, it's a product that evolves with your needs in real-time.

Yes. Email [email protected] and it reaches the person who writes the code, because that is one person. Requests that fit where Aera is going tend to land in weeks rather than quarters. The changelog shows what shipped and when, so you can check that claim before you believe it.