01 / The short version
What happened
LangSmith has added browser capabilities to Managed Deep Agents, allowing agents to interact with websites without an API. The feature includes a remote browser sandbox and tools built on Stagehand, Browserbase's open source SDK. LangSmith claims that on 100 WebVoyager tasks, these tools passed 90% of tasks for $1.98 total, which they state was the highest pass rate among seven tested toolsets. Browserbase's MCP server reportedly passed 89% at nine times the cost. The system supports human handoff for logins and allows saved browser steps to be replayed as Playwright scripts.
See the exact references02 / Key takeaways
What you need to know
- 01
Managed Deep Agents now provides a managed browser sandbox for web-based agent tasks.
- 02
LangSmith claims a 90% pass rate on 100 WebVoyager tasks for $1.98 using Stagehand-based tools.
- 03
Browserbase's MCP server reportedly passed 89% of the same tasks at 9x the cost.
Keep in perspective
What to watch for
Performance and cost claims are attributed to LangSmith and have not been independently validated.
Go to the source
Exact references
These are the original pages used for this brief. Publisher claims are not independent evaluations.
01Primary source · LangChainRead the original announcementhttps://www.langchain.com/blog/browser-use-in-managed-deep-agentsAI-generated from the linked source. It can miss context; verify consequential details in the original. How the radar works