A

Browser Agents & MCP

Curated resources and recent cited research for browser-agent workflows

Durable, reviewed resources plus recent cited reports. Curated resources are reviewed for browser-agent workflows; recent reports are public, no more than 30 days old, and cite at least two distinct web hosts.

15 curated workflow resources meet the content and source-depth standard.

Choose the smallest browser layer that can finish the job

Start with the least privileged route, then escalate only when the task requires more browser context.

1. Public-web researchUse SearchAgentSky when the sources are public. It searches live sources, cites them, and leaves a result you can share.
2. Browser-rendered public pagesUse a small browser or MCP runtime when JavaScript-rendered pages need browser execution but no private account is involved.
3. Signed-in browser actionsEscalate to Unchained when the job needs a dedicated persistent Chrome workspace, a login, or multi-step browser actions with local stop control.

Choose the right browser layer

We ran 12 local tasks and 20 real pages through HTTP fetch and headless Chrome. The agent adapter is not yet configured, so its row stays pending. Use the benchmark to decide when a page needs browser execution.

Read the benchmark-backed decision guide →

Start here: browser-agent workflow resources

Durable, reviewed resources. The recent reports below are a separate, time-bound feed.

Choose your approach

Build and integrate

Measure and compare

Model landscape

Ask SearchAgentSky about your workflow

Describe what you need to automate. Sky will compare live sources and return a cited answer.

No new trusted reports this month. The curated resources above are the durable guide.