Gemini API boosts agents
Google’s Gemini API got an update to enable seamless web search plus system integration for agents, opening up use cases like real‑time competitor pricing and CRM‑grounded outreach. The change is being framed as a practical enabler for agentic enterprise automation. (x.com)
Developers can now combine function calling with Gemini’s built‑in tools (Google Search, Maps, etc.) in a single API request, removing the prior need to orchestrate separate calls between search and custom functions. (blog.google) The Agent Development Kit (ADK) exposes a google_web_search tool and corresponding SDKs (Python, TypeScript, Go, Java) so code‑first agents can run web searches as part of agent toolchains; the ADK docs note the google_search tool is currently compatible with Gemini 2 model families. (google.github.io) Google’s Deep Research agent, powered by Gemini 3.1 Pro, is available in preview and supports background execution for long‑running multi‑step research tasks that can take several minutes to complete. (ai.google.dev) Gemini Enterprise features agent management and governance capabilities and now promotes interoperability via the Agent2Agent (A2A) protocol while adding observability controls in public preview (release notes dated March 12, 2026). (cloud.google.com) Third‑party coverage and vendor guides say the unified tool call reduces orchestration latency and simplifies workflows for real‑time tasks like competitor‑price scraping and CRM‑grounded outreach, but vendors also highlight the need to connect agents to live enterprise systems (Salesforce, SAP, Snowflake) for production readiness. (testingcatalog.com) Public pricing summaries published in early 2026 list Gemini API ranges roughly from $0.10–$4.00 per million input tokens and $0.40–$18.00 per million output tokens across model tiers, adding a concrete cost consideration for high‑frequency agent workflows. (blog.laozhang.ai)
