Tools & agents
AI tools and agents that search, write, code and act. What they do, who can use them and where human oversight matters.
AI news, explained simply
Today
Amazon announces World Product Graph to index all online products
Amazon plans a database of every product sold online, even those it doesn’t carry, to keep shoppers on its site.
AWS adds OpenFold3 research preview to SageMaker JumpStart
AWS users can deploy the open-source structure-prediction model on managed SageMaker endpoints, which process jobs asynchronously and can scale to zero when idle.
SageMaker Unified Studio lets admins tailor project navigation
AWS says administrators can show project members different tools based on their designation, with one layout applying across every project in a domain.
Cloudflare releases Kitesurf browser as open‑source project
The stateless AI‑agent browser Kitesurf is now on GitHub under Apache 2.0, letting developers run and modify it on their own Cloudflare accounts.
Amazon Bedrock AgentCore adds OAuth 2.0 discovery support
Developers can now configure Bedrock AgentCore with OAuth 2.0 discovery URLs, expanding beyond OpenID Connect.
Anthropic launches Claude for Manufacturing plugin for factory software
The new plugin lets Claude access Autodesk Fusion, Oracle NetSuite, Microsoft 365 and Infor to automate quoting, quality reports and supply‑chain tasks, with human review required.
AWS GovCloud adds Grok 4.7 to Kiro IDE and CLI
SpaceXSI's Grok 4.7 model, with a 500 K token window, is now usable in AWS GovCloud (US) via the Kiro development environment.
ElevenLabs rolls out preview of ElevenAI meeting and dictation app
The voice‑AI company launched ElevenAI in preview, letting enterprise teams sign up for early access to combined note‑taking, dictation and agent tools.
Devin launches private web app builder
Devin now lets teams describe a tool and have it built as a private web app on Devin’s cloud machine, with a stable URL and per‑use billing.
81% of software teams use AI‑generated code they don’t fully understand
TestMu AI’s 2026 report finds most developers rely on AI code but few verify it, leading to more production bugs.
Google open‑sources Ambient Quality Agent for AI production debugging
AQuA, Google’s new open‑source tool, scans live AI agent sessions to spot silent failures and point developers to the offending code.
FDA clears Bracco’s UPTUNITY Aneurysm AI tool for brain scans
Bracco and Avicenna.AI received 510(k) clearance for UPTUNITY | Aneurysm, software that flags brain aneurysms over 3 mm on CT angiograms.
New framework guardscope checks LLM agent control coverage
guardscope lets developers verify which action paths runtime controls actually inspect in LLM agents.
ArcticSwarm boosts multi‑agent web search accuracy by delaying early consensus
The new ArcticSwarm architecture separates evidence gathering from integration, raising BrowseComp scores to 82.6% with Qwen 3.5‑27B.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
Yesterday
FAA expands AI tool to help manage US air traffic nationwide
The FAA will roll out its $875 million SMART system to all 22 air‑traffic control centers, aiming to cut delays and fuel use.
CopilotKit launches OpenBot 1.0, a self‑hosted AI coworker
CopilotKit released OpenBot 1.0 under an MIT license, letting companies run their own AI agents with policy checks and audit trails.
Anthropic adds Claude AI to Google Meet calls
Claude Meetings lets the AI sit in on video calls, take notes and draft tasks for teams on Anthropic's paid plans.
Amazon MWAA Serverless adds DuckDB operators
AWS now lets MWAA Serverless workflows run DuckDB tasks, letting users query S3 data with in‑memory SQL without extra credential setup.
Google tests Agent Anomaly Detection on five public benchmarks
In a private‑preview test, the tool flagged 87% of harmful actions but raised a few false alarms on normal office tasks.
AWS adds DevOps Agent to EC2 Image Builder console
The new integration lets users troubleshoot up to 50 failed image pipelines and create resources via chat in the console.
AWS makes next‑gen AgentCore runtime the default in many regions
The new AgentCore runtime starts sessions faster, uses less memory and can scale to zero when idle, now default in 20 AWS regions.
RemotePC adds AI assistant integration for remote computer tasks
RemotePC now works with ChatGPT, Claude and other AI assistants, letting users check status, grab screenshots and schedule restarts via natural language.
AWS adds faster, larger, cheaper AgentCore runtime for AI agents
Amazon Bedrock AgentCore now starts sessions up to 15× faster, supports 4× more compute, and reduces memory use by up to 68%, with pay‑as‑you‑go pricing.
Google DeepMind launches Gemini 4 Hydrogen model
Gemini 4 Hydrogen is now available to all users, offering faster code and agentic tasks at $1.50 per million input tokens.
Study finds AI chatbots sometimes link shoppers to fake retail sites
Google Gemini, OpenAI ChatGPT and Perplexity were tested on 151 brands; risky site links appeared in 4‑15% of answers.
Game developers’ AI use rises to 40% in 2026 survey
The GDC Festival of Gaming survey finds AI tools now part of the job for 40% of developers, up from 36% a year earlier.
Anthropic adds self‑hosted Claude Managed Agents
Anthropic now lets customers run the full Managed Agents service on their own servers, storing data locally and choosing any model provider.
Replit Agent adds buying power and Google Ads for small businesses
Replit says its AI coding assistant can now purchase supplies via Link and Visa and run Google Ads, giving founders more automated tools to grow.
LTM adds AI‑driven quantum error mitigation to IBM Quantum Functions catalog
LTM’s AI‑QEM tool is now listed in IBM’s Quantum Functions service, making the firm the first global system integrator to publish a function there.
Wirex launches Navoka AI agent that pays from its own bank account
Wirex unveiled Navoka, an AI payment assistant with a dedicated IBAN, debit card and crypto wallet, available by invitation only.
Amazon Bedrock adds fine‑grained ACL control to Managed Knowledge Base
AWS blog shows how to limit each agent’s search to documents a user is allowed to see using ACL‑aware knowledge bases and a gateway interceptor.
AWS demonstrates cross‑org agent integration using A2A protocol
AWS blog walks through a reference setup that lets an enterprise LangChain orchestrator talk to a SaaS compliance agent across AWS accounts via the open A2A protocol.
Surescripts and athenahealth add prior‑authorization automation for 200 drugs
The two companies say the new AI‑driven tool now covers almost all prescribers on athenaOne, aiming to speed up medication approvals.
NiCE unveils next‑gen NiCE Cognigy platform for enterprise AI agents
NiCE says its new Cognigy tools let large firms build and run AI agents faster, with reusable skills and full decision visibility.
OpenAI pledges $2 million for AI‑children safety research
OpenAI says it will fund work on tools to assess AI risks and benefits for kids.
Azure Logic Apps Standard adds local CLI preview
Microsoft released a preview of a command‑line tool that lets developers create, run and test Logic Apps Standard workflows locally, without VS Code.
Tuya releases open‑source Skill for Meta's Muse to control smart devices
Tuya's new Skill lets Muse users manage any Tuya‑connected device via natural language using a single cloud API key.
Wednesday, November 11
Amazon Bedrock adds standalone Web Search API
Developers can now call Amazon's web index directly with a single API call, no gateway or extra resources needed.
Google pilots Tabletop AI tool for group planning
Google Labs is testing Tabletop, an AI‑assisted space where groups can create itineraries, polls and more, currently limited to U.S. users on a waitlist.
