LlamaIndex
LlamaIndex builds document parsing and extraction software plus open-source tools for AI applications, helping teams turn enterprise documents into usable data for AI agents and workflows.

Open-source library that lets any LLM drive a real browser — the backbone of many web agents
Browser Use is a platform for running AI agents that operate a web browser, paired with a managed remote browser service for teams that already have their own agents. It targets AI startups, automation teams, and developers who need to complete tasks on sites that block or complicate ordinary automation.
The platform splits into Browser Use Agents and Browser Infrastructure. Agents are designed to finish a task end to end, while the infrastructure product supplies a remote browser that your own agent controls. Both connect through an SDK, a REST API, or the Chrome DevTools Protocol, and the company describes setup as pasting a connection URL. That range of entry points matters because agent frameworks differ in how they attach to a browser, and CDP support keeps the option open for teams already using Playwright or similar tooling.
Browser Use publishes its own benchmarks. On a 106-task Internal Bench Hard, the company reports 82% task success at 17 cents per solved task, compared with 62% at $3.40 for Opus 5, 59% at $1.55 for Sonnet 5, and 59% at $2.20 for Gemini 3.1 Pro. On a separate 71-site protected-site benchmark, Browser Use reports reaching 81% of sites, ahead of Anchor at 77%, Kernel at 67%, Steel at 47%, and Browserbase at 42%. These are vendor-run tests with disclosed setups, so they are useful as directional evidence rather than independent validation.
The pricing page frames model selection as a quality-versus-price decision. Recorded results range from GPT-6 Astra at medium, with a quality score of 80.6 and $8.867 per task, down to GPT-5.6 Luna at default, with 16.1 and $0.056 per task. DeepSeek V4.1 Flash settings land between roughly $0.118 and $0.263 per task with quality scores from 29.0 to 47.9. The spread is wide enough that model choice, not just task volume, drives the bill. Browser Use notes that recorded agent costs exclude judge, browser hosting, runner, and unrecorded calls, and that these are not Cloud prices.
Billing is pay-as-you-go with no subscription, and credits do not expire. The minimum top-up is $5, which unlocks every model, and eligible Google, GitHub, or Microsoft signups receive $1 in free credits once without a card. Browser time runs $0.02 per hour, rounded up to whole minutes with a one-minute minimum. Residential proxies are on by default at $5 per GB; switching to a direct or your own proxy drops that to $0.20 per GB. When you use Browser Use's model access, the company adds a 20% fee on published token rates, and browser time and traffic are billed separately. Concurrency begins at 10 sessions and scales automatically with lifetime spend, reaching 50 sessions at $200, 250 at $1,000, 500 at $5,000, and 1,000 at $25,000.
Browser Use describes itself as one of GitHub's most-starred projects, with an MIT license and the option to run in the cloud or on your own machine. The site links to documentation, a browser harness, custom model support, proxy configuration, and a changelog, and the company publishes guides on agent frameworks and browser infrastructure. A case study from Hermes claims 48–66% fewer task tokens after making Browser Use its default browser tool. The open-source core gives teams a path to self-hosting, while the hosted service handles the infrastructure work.
The product is aimed at teams building agents that must interact with real websites rather than clean APIs. Protected sites, login flows, and multi-step tasks are the stated focus, and the benchmark framing suggests the company sees reliability on those sites as its main differentiator. The tradeoff is that costs accumulate across model tokens, browser time, traffic, and proxies, so teams should model usage against the published rates before committing to a workload.
Pay-as-you-go with no subscription; credits never expire. Minimum top-up is $5, and new eligible signups get $1 in free credits once. Browser time is billed at $0.02 per hour, rounded up to whole minutes with a one-minute minimum. Residential proxies cost $5 per GB and are on by default; direct or your own proxy costs $0.20 per GB. Browser Use adds a 20% fee on published model token rates, and browser time and traffic are extra. Concurrency starts at 10 sessions and scales with lifetime spend: $200 unlocks 50, $1,000 unlocks 250, $5,000 unlocks 500, and $25,000 unlocks 1,000. Model prices and benchmark costs are listed per task and per million tokens.
Features, pricing and availability may change. Always verify details on the official website. Maintained by steven · Last updated Oct 11, 2026 →
LlamaIndex builds document parsing and extraction software plus open-source tools for AI applications, helping teams turn enterprise documents into usable data for AI agents and workflows.
Flowith is an agentic AI workspace that combines a whiteboard-style canvas, access to 40+ image and video models, and mini apps for design, e-commerce, and social content.
Fine-tune open models faster with far less memory
Groq is an AI inference cloud for developers and enterprises that need low-latency model responses at scale, delivered through custom LPU hardware and a tiered platform spanning bare metal, inference, and enterprise controls.