Creativity · Agent Protocol
Selenium for AI Agents
Selenium predates modern web-agent tooling by two decades but remains relevant. Enterprise QA teams have deep Selenium infrastructure; some browsers (older IE-like environments, specialized embedded browsers) only support WebDriver, not CDP. AI agents in those contexts often use Selenium with an LLM steering layer bolted on.
Protocol facts
- Sponsor
- Selenium project (OSS foundation)
- Status
- stable
- Spec
- https://www.selenium.dev/
- Interop with
- WebDriver BiDi, Selenium Grid, LangChain tools
Frequently asked questions
Should I pick Selenium over Playwright for a new agent?
Usually no. Playwright is faster, has better auto-waiting, and a nicer API. Pick Selenium only if you need WebDriver BiDi cross-vendor support, an existing Selenium Grid, or specific legacy browser compatibility.
What about Selenium + LLM?
The pattern is the same as with Playwright: the LLM decides actions, Selenium executes them. LangChain and a few OSS projects provide Selenium tool wrappers for agents.
What's WebDriver BiDi?
The new bidirectional WebDriver protocol that closes the feature gap with CDP (events, network interception). Once widely adopted, Selenium-based agents get near-Playwright capability across all major browsers.
Sources
- Selenium docs — accessed 2026-04-20
- WebDriver BiDi spec — accessed 2026-04-20