YOUR BRAND · YOUR CUSTOMERS · YOUR AI BUSINESS

Sell AI employees under your own brand.

AI Agent News Today

Friday, November 20, 2026

LangChain launches LangSmith Tool Gateway (public beta)

What changed: LangChain put LangSmith Tool Gateway into public beta on November 19, 2026; the gateway centralizes credentials, caller identity, access policies, approval flows and audit traces for every agent tool call and works with MCP-compatible clients (LangChain, Deep Agents, Claude Code, Cursor, etc.).

Why it matters: This targets the practical blocker most teams hit when agents move from prompts to scheduled or event-driven work: who the agent acts as and which tools it can call. With a managed kit and hosted connectors for Google, Microsoft, Salesforce, GitHub and more, teams can speed safe production rollouts without hand-building auth per tool.

Try/watch: If you run agents, enroll in the LangSmith public beta, build a minimal kit with 1–2 trusted connectors, and verify deny/approval policies and audit logs under real error conditions; watch for an independent security evaluation and the GA timeline before putting high‑risk actions behind agents.

Databricks opens Genie ZeroOps to public beta (autonomous pipeline agent)

What changed: Databricks announced Genie ZeroOps in public beta on November 19, 2026; the agent continuously monitors jobs and tables, diagnoses failures using Unity Catalog metadata and lineage, runs dry‑run tests, and proposes fixes as PRs while requiring human approval before changing production.

Why it matters: Data teams can reduce hours spent triaging broken pipelines because ZeroOps ships with direct access to workspace lineage, run history and repository integration, letting it surface prioritized root causes and a one‑click PR to fix common issues. The product is built to run inside your Databricks security perimeter and respects workspace permissions.

Try/watch: Enable ZeroOps in a sandbox workspace, test its detection-to-PR flow on noncritical pipelines, and add custom instructions to tune noisy alerts; watch how it handles complex cross-team fixes, and guard any agent-run attempts to modify production without explicit approval.

Microsoft Foundry gives enterprise agents their own Microsoft 365 identity

What changed: Microsoft detailed Microsoft Foundry updates on November 19, 2026 that let organizations provision agents as named Microsoft 365 members with their own Entra (agent) identities, email, calendar, OneDrive and an accountable manager, plus blueprints for company‑wide or team‑level agents.

Why it matters: Making agents first‑class members of the org separates capability from access: you can give an agent the ability to act without borrowing a user’s credentials, which simplifies audit, permissioning and compliance for agents that must operate across teams and group spaces.

Try/watch: Pilot a Foundry blueprint that assigns an agent a manager and narrow permission scope, test group‑chat and shared-drive scenarios, and audit every agent action; monitor how Entra tooling and existing enterprise governance scale as you add dozens of persistent agent identities.

More News

YOUR BRAND · YOUR CUSTOMERS · YOUR AI BUSINESS

Sell AI employees under your own brand.

Give businesses AI teams that own ongoing roles. Set your pricing, delegate routine coordination and build a recurring service.