AI Agent News: The Biggest Launches of November 2026

Last Updated: November 12, 2026

Summary: The biggest AI agent news of the past month is that agents moved from chat windows into everyday work channels and enterprise back offices. Anthropic released Claude Fable 5.2 on November 9 with agent orchestration built in, IBM and Microsoft launched the Imagine Enterprise agent platform on November 10, OpenAI cut agent model prices with GPT-6.1 Terra, and security companies spent heavily to control what agents can access. The next dates to watch are Microsoft Ignite (November 17 to 20) and the November 30 shutdown of OpenAI's Agent Builder.

AI agents are software that take a goal, plan the steps and carry them out with tools, rather than just answering a question. This living tracker lists the most important agent news of the last 30 to 45 days, then each company's agent products, key trends and what to watch next.

AI agents in November 2026 are cheaper to run, easier to reach and harder for security teams to ignore. Three shifts stand out.

The takeaway: agent costs per task are falling while governance work is growing, so budget for both. For the basics, see our explainer on what AI agents are.

The most important AI agent launches and deals from the last month, newest first, with what each means for businesses.

IBM and Microsoft announced Imagine Enterprise, a platform for building, running and governing AI agents inside industry-specific business processes, according to IBM's newsroom. Imagine Enterprise draws on more than 6,000 pre-built agents and assets from IBM Consulting, runs on Microsoft's cloud and security foundation, and starts with financial services, insurance and retail. Why it matters: regulated companies stuck at the pilot stage get a packaged option with compliance controls and engineers from both vendors.

Meta released Muse Ember, an entry-level model available through Meta's Model API at $0.10 per million input tokens and $0.50 per million output tokens, SiliconANGLE reported. Meta says Muse Ember is optimized to handle smaller tasks assigned by a larger frontier model, and it offers a 1 million token context window. Why it matters: setups where an expensive model plans and a cheap model executes are becoming the default, and another low-price option pushes per-task costs down.

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, the same underlying model with different safeguards. Fable 5.2 can work as an Advisor that smaller Claude models consult on harder problems, or as an Orchestrator that delegates simpler tasks to them. Anthropic says the model coordinated more than 1,000 subagents for eight days in one experiment. API pricing is $8 per million input tokens and $40 per million output tokens, and Mythos 5.2 is limited to vetted organizations. Why it matters: orchestration is now a built-in model feature rather than something every team has to engineer itself. See our guide to Claude AI agents.

Amazon began rolling out texting for Alexa for Shopping to select US customers by invitation, with plans to reach millions more in the coming weeks, according to Amazon's announcement and TechCrunch. By text, the agent can search, build a cart, buy, track prices and check order status. TechCrunch noted the launch comes after Amazon blocked Meta's Muse agent from querying its site in September. Why it matters: large retailers are running their own shopping agents rather than letting outside agents shop their sites, which shapes how brands plan for agent commerce.

Palo Alto Networks is in talks to buy Clutch Security, a startup that secures AI agents' access to company systems, for about $500 million in cash plus milestone payouts, TechCrunch reported, citing a person familiar with the talks. The Information first reported the talks. The deal is unconfirmed, and neither company responded to requests for comment. TechCrunch put the price at about 50 times Clutch's annual recurring revenue, which is approaching $10 million. Why it matters: controlling what agents can log into is becoming its own security category, and buyers are paying high multiples for it.

Perplexity launched Teammates, personal agents for long-running work that keep track of project goals, task history and shared files across runs. Teammates can join Slack channels, where colleagues assign tasks or ask for updates, and are available to Perplexity Max subscribers, starting on the web. Why it matters: agents that persist for days inside team chat need clear owners and review steps.

CrowdStrike announced an agreement to acquire Fiddler, which provides AI agent observability and explainability, and plans to sell it as Falcon Observe, according to CrowdStrike's press release. The price was not disclosed, and the deal is expected to close in CrowdStrike's fourth quarter of fiscal 2027. "Fiddler is the flight recorder for every agent: the observability to explain exactly what an agent did and why," said CrowdStrike CEO George Kurtz. Why it matters: auditors will ask why an agent took an action, and security platforms are building that record in.

OpenAI began a beta that lets ChatGPT Pro subscribers in the US text their Dot agent over iMessage or RCS, Digital Trends reported. Users can assign tasks or ask the agent to monitor something, and the agent can text back proactively when it has an update. Why it matters: always-on agents reachable by text raise expectations for how fast AI tools respond.

OpenAI launched GPT-6.1 Terra, which it calls "a fast, efficient model that combines detailed visual reasoning and precise computer use," InfoWorld reported. API pricing is $1 per million input tokens and $5 per million output tokens, half of GPT-6.1 Sol's $2 and $10. On the OSWorld 2.1 computer-use benchmark, Terra scored 74.2% at $1.19 per task. Why it matters: agents that operate software through a screen are getting cheaper per task, which makes back-office automation pilots easier to justify.

Gartner's top 10 strategic technology trends for 2027 include AI agent identity and AI coding agent harnesses, Network World reported. Gartner predicts that by 2028, 25% of enterprise security breaches will trace back to AI agent abuse, and that by late 2026 autonomous coding agents will generate 20% of all public commits. Why it matters: giving every agent a managed identity, with its own permissions and audit trail, is moving from best practice to an expected control.

Google introduced a unified Gemini agent, starting with businesses before consumers, at a Google Cloud event, TechCrunch reported. Google Cloud CEO Thomas Kurian said the agent can be given "objectives, not just instructions." The agent can hand work to subagents, connects to Workspace, Microsoft 365, Slack, Jira and any MCP server, and has its own Workspace account so its actions are logged under the agent's name. Why it matters: an agent with its own email address and audit trail changes how IT teams handle access and accountability. See our breakdown of Google AI agents.

The main AI agent platforms in late 2026 come from OpenAI, Anthropic, Google, Microsoft, Amazon and Salesforce, with Meta and Perplexity pushing consumer and prosumer agents. For a deeper comparison of builder tools, see our guide to AI agent platforms.

The pattern is similar across vendors: a hosted runtime, a control plane to govern agents and pricing that mostly follows token use.

The key AI agent trends are wider enterprise scaling, falling cost per task, agent security consolidation and agent identity.

McKinsey's State of AI 2026 survey, published August 25, 2026, found that about 20% of respondents say their organizations are scaling AI agents across the enterprise. Among organizations with more than $1 billion in annual revenue, 40% report scaling agents in at least one function, up from 27% a year earlier, while smaller organizations were flat at 22%. The survey drew 1,719 participants in 97 countries. More numbers: AI agents statistics.

Budget agent models keep getting cheaper, and the gap between frontier and budget models is now 80 times on list prices. That makes model choice per step matter more than the headline model.

A worked example at list prices: an agent step that reads 100,000 tokens and writes 10,000 tokens costs $1.20 on Claude Fable 5.2 ($0.80 plus $0.40), $0.15 on GPT-6.1 Terra ($0.10 plus $0.05) and $0.015 on Muse Ember ($0.01 plus $0.005). Route routine steps to cheaper models and save the frontier model for planning, but test quality on your own tasks first.

Security vendors are buying startups that watch and control agents. TechCrunch's report on the Clutch talks also noted that Cyera bought Oasis Security, which manages AI agents' system access, for $1 billion, and that Cisco bought Astrix for roughly $400 million according to Calcalist. Separately, SiliconANGLE reported that Cisco's 2026 Relentless Defense report found only 36% of Australian organizations apply consistent controls to AI agents and nonhuman identities.

Google's Gemini Agent has its own Workspace account, Agent 365 registers and governs agents, and Gartner, as reported by Network World, warns that "if organizations don't resolve these identity challenges, over half of their enterprise AI initiatives will halt entirely."

The next big AI agent moments are Microsoft Ignite, the end of OpenAI's Agent Builder and Salesforce's long-horizon agent launch, all due before December.

Sameer's take: My advice for a small business new to AI is to start by automating one boring task.

If your company is planning its first agent project, our step-by-step guide on how to build AI agents covers the design choices and costs.

What is the latest AI agent news?
As of November 12, 2026, the latest AI agent news is IBM and Microsoft's Imagine Enterprise platform (November 10), plus Anthropic's Claude Fable 5.2 and Meta's low-cost Muse Ember model (both November 9).

Which companies make the leading AI agents in 2026?
The leading AI agent platforms come from OpenAI (Agents API), Anthropic (Claude Managed Agents), Google (Gemini Enterprise Agent Platform), Microsoft (Agent 365), Amazon (Bedrock AgentCore) and Salesforce (Agentforce).

Are AI agents safe for business use?
AI agents can be used safely with scoped permissions, a managed identity for each agent and logs of every action. Gartner predicts that by 2028, 25% of enterprise security breaches will trace back to AI agent abuse.

How much do AI agent platforms cost?
Most agent platforms charge for model tokens, sometimes plus a usage fee. OpenAI's Agents API adds no fees beyond tokens and tools, Claude Managed Agents adds $0.08 per session hour, and Agent 365 costs $15 per user per month standalone, Thurrott reported.

What AI agent events are coming up?
Microsoft Ignite runs November 17 to 20, 2026, in San Francisco with sessions on Agent 365 and agent security. OpenAI's Agent Builder shuts down on November 30, 2026.

The AI agent news of the past month points one way: agents are getting cheaper per task, showing up in text messages and team chat, and taking on longer jobs, while security and identity controls race to keep up. For most businesses, the right move is to pilot one well-scoped agent with its own identity and permissions.

⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.

Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.