BrowserStack
Also known as: browserstack.com, BrowserStack AI, BrowserStack Test Companion, BrowserStack MCP Server
BrowserStack runs a cloud of more than 3,500 real desktop browsers and mobile devices for manual and automated testing, and now ships a set of AI agents across the test lifecycle. Agents turn requirements into test cases, convert them into low code automation, heal broken locators while builds run, analyze failures, select which tests to run, review visual changes and find and fix accessibility issues. Test Companion brings the same work into the IDE, and an MCP server opens the device cloud to Claude, Cursor and VS Code.
BrowserStack sells testing infrastructure for web and mobile teams: Live and App Live for manual testing on real browsers and devices, Automate and App Automate for running Selenium, Playwright, Cypress and Appium suites in its cloud, Percy for visual testing, accessibility testing, test management and low code automation. The real device cloud spans more than 3,500 desktop browsers and devices across 21 data centers in 13 locations in the US, Europe and Asia Pacific and runs about 1 billion tests a year. More than 50,000 customers in 135 countries use it, among them Google, Amazon, Atlassian, Stripe and LinkedIn.
AI agents are built into those products. The Test Case Generator turns requirement documents into test cases, the Low Code Authoring agent turns test cases or plain language prompts into automated tests that run on real browsers, the Self Healing agent repairs broken UI locators at runtime without code changes, the Test Failure Analysis agent reads run logs to find the root cause and suggest a fix, and a Test Selection agent picks which tests to run. The Visual Review agent highlights only meaningful visual changes in Percy, the A11y Issue Detection and Remediation agents find and fix WCAG problems that rule based checks miss, and a Test Deduplication agent removes duplicate cases.
Test Companion is the agentic option for engineers: inside the IDE it generates test cases, writes automation scripts for Playwright, Cypress, Selenium or Appium, covers biometric and passcode flows on real phones, explores for edge cases and fixes failing tests. The BrowserStack MCP Server, local or remote over OAuth, lets assistants such as Claude, Cursor and VS Code run manual and automated tests on real devices, scan for accessibility issues, create test cases and runs and call the AI agents, and REST APIs for Automate and Test Management open the same data to a team's own scripts.
Each product is sold on its own plans. Live starts at $12.50 a month billed annually for a freelancer, Automate at $99 a month, App Automate at $199 and Percy at $199, with free tiers for Percy, App Percy and accessibility scans. Self Healing, Test Failure Analysis and Test Selection come with the Pro automation plans, the Visual Review agent with Percy, accessibility agents with Accessibility Ultimate, and the Test Case Generator with the Team Ultimate pack at $375 a month. BrowserStack is SOC 2 Type 2 compliant and offers SSO, IP allow listing and access control as Enterprise add ons.
It fits QA and engineering teams that already run tests on BrowserStack and want AI to write, maintain and triage those tests on the same device cloud. It fits less well for buyers who want to choose the model behind each agent, pin test data to one region, or approve each automated fix before it applies.
Vendor details
Canonical URL
https://www.browserstack.com
Category
Agent infrastructure
Subcategory
Real browser and device testing cloud with AI agents that write, run, heal and triage tests
Funding status
Private company, founded in 2011 by Ritesh Arora and Nakul Aggarwal and backed by Accel, BOND and Insight Partners; its 2021 Series B of $200 million valued it at $4 billion.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Agents and Test Companion work with Playwright, Cypress, Selenium and Appium, CI and CD tools such as Jenkins and GitHub, and Jira and Slack. The BrowserStack MCP Server connects Claude, Cursor and VS Code to the device cloud, accessibility scans, test management and the AI agents, locally or remotely over OAuth, and REST APIs for Automate and Test Management let teams pull build and session data and create or update test cases from their own tools.
In practice
A retail app team uploads its checkout requirements to the Test Case Generator, gets 40 test cases back, and has the Low Code Authoring agent turn the top 15 into automated tests that run on real iPhones and Android phones before each release.
A nightly Playwright suite on Automate breaks after a front end redesign renames buttons. The Self Healing agent repairs the locators during the run so the build passes, and the Test Failure Analysis agent explains the two real failures that remain.
An engineer asks Claude, connected through the BrowserStack MCP Server, to run the signup flow on Safari and three Android versions and scan the page for WCAG issues, then fixes the contrast problems the A11y agent flags.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
10.0 / 14 capabilities · 71%
| Integrations & Tool Calling | Full |
|---|---|
|
The agents work directly in the customer's testing estate: they create projects, test cases and runs in test management, update results, rewrite automation for Playwright, Cypress, Selenium and Appium suites and run them on BrowserStack's devices. Test Companion connects into CI and CD pipelines such as Jenkins and GitHub and into Jira and Slack, and the MCP server lets outside assistants call the same test management and device actions. Sourcebrowserstack.com/docs/browserstack-mcp-server/overviewread 2026-11-18 |
|
| Workflow Orchestration | Full |
|
The agents hand work along the test lifecycle: requirements become test cases, test cases become low code automation, runs are healed as they execute, failures are analyzed and the next run picks only the tests that matter. Test Companion covers the same chain from the IDE, planning, authoring, executing, debugging, maintaining and fixing tests, with specialized agents taking the failure analysis and healing steps inside one session. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Knowledge Grounding & RAG | Partial |
|
Agents ground their work in the material handed to them for each job: product requirement documents for the Test Case Generator, with ten uploads per user each month on the Team Ultimate pack, run logs for failure analysis, screenshots for visual review and page structure for accessibility checks. Each agent reads what it is given rather than searching a standing knowledge base of the team's documents. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Human Oversight & Guardrails | Partial |
|
Visual Review narrows a build's screenshot changes to the meaningful ones for a person to accept or reject, and Test Failure Analysis suggests a fix rather than applying it. Self Healing, by contrast, repairs locators during the run with no step where a person approves the change first. Customers can opt in or out of the AI features at any time, and their data is used only to produce output for them. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Security, Identity & Governance | Full |
|
BrowserStack is SOC 2 Type 2 compliant across its data centers and platform and has completed a CSA STAR Level 1 self assessment, with devices torn down after each session. Enterprise plans add single sign on, IP allow listing and access controls over who on a team can use which products, and the MCP server's remote option signs users in over OAuth. Sourcebrowserstack.com/securityread 2026-11-18 |
|
| Observability & Auditability | Partial |
|
Test runs keep the logs the Test Failure Analysis agent reads to explain why a build failed, and Visual Review shows exactly which screenshots changed between builds. No separate record lists what the agents themselves changed, such as each locator the Self Healing agent swapped during a build, so teams trace an agent's effect through each run's own logs and outcomes. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Memory & State Persistence | Not documented |
|
Test cases, automation scripts and run history stay in the customer's test management and automation projects, and the agents work on those records each time they run. They keep no memory of their own, with a scope and retention, of preferences or past fixes that carries from one session to the next, and devices are torn down after every session. History lives in the test records. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Deployment & Data Residency | Partial |
|
Tests run across 21 data centers in 13 locations, seven in the US, three in Europe (Dublin, Frankfurt and the Netherlands) and two in Asia Pacific (Mumbai and Sydney), and sessions connect to the closest one automatically. A Custom Device Lab offers tailored environments, but customers are not offered a setting to keep their tests or data in one chosen region. Sourcebrowserstack.com/data-centersread 2026-11-18 |
|
| Prebuilt Agents / Templates / Packs | Full |
|
A named set of ready agents ships inside the products, each doing its own job: Test Case Generator, Low Code Authoring, Self Healing, Test Failure Analysis, Test Selection, Visual Review, A11y Issue Detection, A11y Issue Remediation and Test Deduplication, plus Proofreader and Form Filler in the Team Ultimate pack. Each switches on with the plan that includes it, with no setup required. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Triggers & Channel Coverage | Full |
|
Several agents work inside runs that CI starts with nobody watching: the Self Healing agent repairs broken locators at runtime, the Test Selection agent decides which tests a build runs, and the Test Failure Analysis agent works through the logs when a run fails. Percy's Visual Review agent fires on each new build's screenshots, and engineers can also call agents on demand from the IDE through Test Companion or from an assistant through the MCP server. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Model Flexibility & Routing | Not documented |
|
Each agent runs on models BrowserStack selects for the job and tunes with its own prompts, with no model or provider named. Customers cannot pick a model for any agent or for Test Companion, and teams that want their own assistant in the loop connect it to the device cloud through the MCP server instead, which changes where they work rather than which model powers BrowserStack's agents. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
REST APIs reach the platform itself. The Automate API at api.browserstack.com returns plan, project, build and session data within set rate limits, and the Test Management API creates, updates and deletes test cases one at a time or in bulk, both signed in with the account's username and access key. The BrowserStack MCP Server runs locally with those credentials or remotely over OAuth, and lets Claude, Cursor and VS Code run manual and automated tests on real devices, scan pages for accessibility, create projects, test cases and runs, update results and call the AI agents. Sourcebrowserstack.com/docs/test-management/api-reference/test-casesread 2026-11-18 |
|
| Testing, Debugging & Optimization | Full |
|
Testing is the product: automated suites run as quality gates on every build, Test Selection chooses the tests a change needs, Test Failure Analysis sorts real failures from noise and points to the root cause, Visual Review flags only meaningful visual changes, and the accessibility agents check pages against WCAG. Teams measure their own apps against expected behavior on real browsers and devices before release. Sourcebrowserstack.com/test-platform/browserstack-airead 2026-11-18 |
|
| Browser / Computer-use | Full |
|
BrowserStack hosts more than 3,500 real desktop browsers and iOS and Android devices, and its agents work them directly: the Low Code Authoring agent builds tests that click and type through the app under test, Self Healing keeps them running when the interface changes, and Test Companion writes tests for real phones down to biometric and passcode logins. Through the MCP server an assistant can drive manual sessions on those same devices. Sourcebrowserstack.com/docs/browserstack-mcp-server/overviewread 2026-11-18 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Live from $12.50 a month billed annually; Automate from $99; App Automate from $199; Percy from $199 with a free tier; Enterprise quoted
Per product plans priced by users, parallel tests or screenshots, with AI agents included in particular tiers
Included quota
Percy Free includes 5,000 screenshots a month and Percy Desktop 10,000. The Test Case Generator in Team Ultimate allows 10 requirement document uploads per user a month.
What is public
BrowserStack prices each product on its own. Live starts at $19 a month or $12.50 billed annually, Automate at $99 a month, App Automate at $199 and Percy at $199, and the Team Ultimate pack of 15 products costs $375 a month. Enterprise plans are quoted.
Billing mechanics
Plans bill monthly or annually per product, and AI agents switch on with the tier that includes them rather than as a separate charge. Enterprise adds single sign on, IP allow listing, access control and priority support.
Cost watchouts
Each agent comes with a specific plan: Self Healing, Test Failure Analysis and Test Selection with the Pro automation tiers, Visual Review with Percy, accessibility agents with Accessibility Ultimate and the Test Case Generator with Team Ultimate.
Variable cost rationale
Plans are flat monthly prices, with metered charges limited to extra Percy screenshots.
Additional watchouts
Prices for Test Management, Low Code Automation and Test Companion are not listed with the other products.
Overage / add-ons
Extra Percy screenshots cost $0.036 each on Desktop and $0.048 on Desktop and Mobile.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free tiers for Percy, App Percy and accessibility scans, and a free trial
Lowest paid plan
Live Freelancer, $19 a month or $12.50 billed annually, with 100 minutes for one user
Commercial notes
New Automate Desktop customers get three free months of Live Team for five users.
Missing data
BrowserStack gives no Enterprise prices and no figures for most Accessibility tiers.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…
- AlphaBitCore — AI control plane for regulated financial firms: one gateway enforces…
Alternatives to BrowserStack
Other products that do the work BrowserStack does, and what each one is.