In 2026, AI software has moved past the per-seat price, and infrastructure providers charge for effort: the compute, the agent hours and the gigabytes your agents use. An agent does more work without adding a person to the account, so the bill now follows what your agents actually do.
That shift puts the billing unit in front of every team running browser agents. Kernel pricing is a plan fee plus per-second usage; Solari bills an hourly browser rate from a monthly credit balance that also pays for sandboxes and desktops. This guide covers Kernel's plans, its meter and its extras, what the same month costs on Solari, and how to compare the two bills for your own fleet.
Why Does the Billing Unit Matter Now?
The billing unit decides which part of your workload you pay for. When teams ran headless browsers on their own servers, the cost was the server, whatever the agent did on it. A browser vendor charges for the browser itself, so the way your agents spend their time (busy, idle, waiting on a model) moves the invoice.
Two vendors can come out close on one workload and far apart on another. A per-second meter with a standby state rewards agents that idle between short bursts. An hourly rate drawn from one credit balance rewards long, busy sessions and teams that run code and desktops next to their browsers.
What Each Kernel Plan Includes
Kernel's plan sets the floor of every invoice and decides which features you can turn on. Kernel says its free plan is $0 a month plus usage with $5 of monthly credits and five concurrent browsers, Hobbyist is $30 a month plus usage with $10 of credits and ten concurrent browsers, and Start-Up is $200 a month plus usage with $50 of credits and 150 concurrent browsers, on its pricing page.
Kernel puts bring-your-own proxies, GPU acceleration and regional browsers on Start-Up and above, and HIPAA on custom Enterprise contracts, its docs say. One feature you need can pick your Kernel plan before usage does.
How Does Kernel Meter Browser Time?
Kernel charges by the second, at a rate set by the kind of browser. Kernel says a headless browser costs $0.0000166667 per second, a headful browser $0.0001333336 and a headful browser with GPU acceleration $0.0008000016, in its pricing docs. Its pricing page quotes usage per GB-second, so browser memory moves the charge as well as time.
The meter runs only while a browser works. Kernel's docs say a browser enters Standby Mode within seconds of its last CDP or Live View activity and stops accruing usage, and that the idle timeout is a deletion ceiling rather than a billing window. An agent that waits on a model between page actions pays Kernel for the actions, not the wait.
What Do the Extras Cost on Each Platform?
Proxy traffic and CAPTCHA solving are the extras that move an agent bill most, and the two vendors price them in opposite ways. Kernel says its free plan includes stealth mode with a captcha solver and proxy, that Hobbyist adds unlimited proxy usage, and that usage carries no charge for idle time or proxies. For a heavily proxied fleet, that bundle is Kernel's strongest line.
Solari prices the same extras per use, so you pay only for the traffic and solves your agents need. Solari charges $6.00 per GB for proxies on Starter and $4.00 per GB on Professional, and $0.01 per captcha solve on Starter and $0.005 on Professional. Both run on stealth sessions, included on every paid plan, and the blockHeavy proxy option stops video and large images before they cross the proxy, so you don't pay for bytes your agent never reads.
Logins and replays round out the extras. Kernel scales managed auth connections, vaults and replay retention with the plan. On Solari, profiles save cookies and site data in Playwright's storageState format, so your agent signs in once, and a session is recorded only when you set the recording flag.
What Does the Same Work Cost on Solari?
Solari swaps the per-second meter for one hourly rate per plan, and the monthly fee comes back as credits that every browser, sandbox and desktop draws from.
Idle time and extras are the rows that change the answer. Kernel stops metering in standby and bundles proxies, while Solari bills per hour while a session is open, so releasing sessions between long model calls keeps your cost tied to work.
Here is what that model does for your bill:
A rate that falls as you grow. Every step up the chart above lowers what an hour of agent work costs you. Enterprise runs browsers at a tenth of Team's rate, so a fleet at internet scale pays a fraction of what a small team pays per hour.
Fees that come back as credits. Starter's $20 fee and Professional's $200 fee each return as the same amount of credits every month. You spend them on any mix of browsers, sandboxes and desktops from one balance.
No surprise overage. When the balance reaches zero, new sandboxes and desktops stop launching until you top up, and there is no end-of-month overage charge. The invoice never runs past the balance you chose.
Room to run in parallel. Professional allows 150 concurrent browsers with sessions up to 24 hours. Solari supports up to 100,000 concurrent browsers on Enterprise, so you can run a whole job list in one pass.
A browser the moment you ask. Solari browsers are ready in 8 ms, so your agent starts work as soon as it requests a session. Short jobs spend their time on the page, not on startup.
The same credits pay for sandboxes that run your agent's code and desktops for computer-use models. Code that already drives a remote browser over the Chrome DevTools Protocol connects to a Solari session through its CDP endpoint, so moving from Kernel needs a new endpoint rather than a rewrite.
How Do You Compare the Two Bills?
The fastest way to compare is to price last month's real work on both models:
Split usage by browser type. Export your Kernel usage and separate headless, headful and GPU seconds, since each runs at its own rate. The split shows which rate actually drives your bill.
Measure idle time. Compare the hours your sessions stayed open with the seconds Kernel metered. A large gap means standby is doing real work for you; a small one means your agents are busy for most of the time they hold a browser.
Price your proxy traffic. Add up the gigabytes your agents sent through proxies and the captchas they hit. On Kernel that traffic is bundled; on Solari it is a per-GB and per-solve line you can shrink with
blockHeavy.Count your other machines. List the code sandboxes and desktops you pay another vendor for. On Solari they draw from the same balance, so the comparison covers your whole agent stack, not just browsers.
Run one job on both. Run a real workflow on a free Solari account and compare the charge with the same job on Kernel. A measured job settles the question faster than a spreadsheet.
Where Is Agent Infrastructure Pricing Going Next?
Agent infrastructure pricing is still settling. Over the next year it moves in four directions:
Billing at the real length of a job. Short sessions will be priced to the fraction of a cent instead of rounded to a minimum, and Solari already carries the unfilled part of a cent onto the next charge on Enterprise.
One balance for every surface. Teams will buy browsers, sandboxes and desktops from one account and budget them as one line, instead of reconciling three vendors.
Limits instead of surprises. Balances, caps and alerts will become standard, so a runaway agent stops at a limit rather than at the end of the month.
Rates that keep falling. Per-hour prices for browsers and VMs will keep dropping as fleets grow, and volume contracts will pull further below list rates.
What stays human is the judgment: which jobs are worth an agent's time, what the budget should be, and whether the invoice bought work that mattered.
Which Pricing Model Fits Your Agents?
Kernel charges a plan fee plus per-second usage, pauses the meter in standby and bundles proxies; Solari bills one hourly rate per plan from a balance that covers browsers, sandboxes and desktops, and prices proxies and captchas per use. Kernel fits agents that idle between short bursts and proxy heavily. Solari fits long, busy sessions, teams that want extras priced line by line, and fleets that run code and desktops next to their browsers.
Kernel is the more established vendor with the larger browser ecosystem; Solari is the younger company and moves faster on speed and price. The introduction to Solari covers the three surfaces behind that single bill.
Get Started With Solari
Start with the busiest workflow you run on Kernel today and run it on Solari to see the charge for the same job side by side. Solari is the fastest and cheapest way to run cloud browsers and VMs for web agents. Get started with a free account and start running your agents today.
