LangChain launches LangSmith Tool Gateway in public beta

LangChain launched LangSmith Tool Gateway in public beta on November 19, adding a managed layer for controlling AI agents’ access to external tools.
The official announcement says teams can register Model Context Protocol servers or use hosted integrations for services including Google, Microsoft, Salesforce, GitHub and Linear. Selected tools are grouped into kits exposed through a single URL, limiting the tools presented to an agent.
Calls can use a person’s credentials or credentials configured for an agent. Administrators can apply deny policies and require human approval before sensitive tool calls run. LangChain directs customers to its team for access to deny policies and the audit log.
With tracing enabled, records include the tool, server, user, outcome and duration, without storing data sent to or returned by the tool. Organization-wide audit logs are an Enterprise feature.
The gateway works with MCP clients including Claude Code and Cursor. Its practical role is to centralize authorization and execution records that teams would otherwise implement separately for each agent; the announcement does not establish an independent security guarantee.
Sources and citations
Revision history
Published .