AI Wiki · Mediocre Security

Products & tools

Agents, frameworks, platforms, chips and services. 68 articles.

ArticleSummaryLinks
AgentOpsAgent-specific tracing and monitoring, proprietary.0
Amazon Bedrock AgentCoreAWS's managed runtime for deploying agents with built-in tools and memory.4
Amazon TorchNeuronNative PyTorch backend for Trainium2 and Trainium3.4
AMD HelionOpen rack reference design for agentic AI shown at the OCP Global Summit on Oct 12 2026: up to 32,768 Zen 6 EPYC cores per rack or EPYC paired with Instinct MI350P GPUs, an integrated token router for agent model routing,…3
Anthropic FlashProveExperimental research and code to verify a response came from the model a provider claims, using cryptographic proofs without exposing private weights.2
Apple Siri AIOpened to developers in the iOS 27.3 betas: apps expose content and actions through App Intents, with multi-step "skills".1
ARTEXOpen-source autonomous penetration-testing agent used in real intrusions; its author pulled the repo and went closed-source in October 2026.32
AutoGenMicrosoft's early multi-agent framework.1
Block DessaVoice model from Block trained entirely on human speech.3
ChatGPTOpenAI's chatbot, launched November 30 2022; the fastest-growing consumer app of its time and the product that started the generative-AI boom.23
Citrix SecurSpacesExpanded Nov 5 2026 to run agents in isolated workspaces with restricted file access, protected credentials and a central control plane, built on SecurAccess ZTNA and NetScaler AI Gateway.1
Claude Agent SDKAnthropic's own production agent-building toolkit.2
Claude CodeAnthropic's agentic coding tool, launched in 2025, that works in the terminal, IDEs and the cloud; one of the most widely used coding agents and a frequent supply-chain target.26
Claude DuetResearch-preview proactive agent app for iOS, Android and claude.ai, designed around Claude Opus 5.5.3
Claude Personal FinanceUS beta launched Nov 5 2026 for Pro, Max and Ultra subscribers.3
Claude SecurityPublic beta expanded to Claude Enterprise on Oct 21 2026: continuous repository scanning, automatic fix tracking, threat models, plus an API and MCP connector.3
Claude Security CenterClaude Enterprise security update: a Security Center for admins and inference hooks, now generally available, that let customers run DLP and policy checks on prompts before they reach the model.1
Cloudflare ClefApache 2.0 vision-capable decision models on Hugging Face and Workers AI, published Oct 1 2026, Jev-API compatible with 64K context.5
CountersignRequires physical-device confirmation that an agent cannot simulate before a spend-limit change.1
CrewAIRole-based "crews" of agents.1
CrowdStrike Falcon GuardianLets authorized admins stop a dangerous agent's execution, block relaunch and restrict its network access.7
DeepSeek Harness (dsh)DeepSeek's open-source Node.js agent runtime where everything is a plugin: model adapters, tool registry, session logging and the agent loop are all swappable.2
GitHub AI Scan and Agentic AutofixAI Scan for pull requests went GA Oct 15 2026, giving advisory findings for languages CodeQL does not cover.2
Google Agent Development Kit (ADK)Google's own production agent-building toolkit.2
Google TorchTPULets PyTorch developers run on TPUs with fewer code changes.3
HackerRank ChakraAI interviewer that went GA after a six-month beta covering 500,000 interviews.1
IBM watsonx OrchestrateIBM announced Oct 27 2026: Discover finds third-party agents, models and MCP servers across Amazon Bedrock, Microsoft Foundry and Google Gemini.3
Intel GPU D270Intel's inference-focused Xe3P PCIe 5.0 card, launched at OCP on Oct 12 2026: up to 480 GB LPDDR5X, 1.3 PFLOPS FP8, 350 W air-cooled.1
JamAI BaseOpen-source, spreadsheet-style UI for building RAG and LLM workflows without code.2
LangChainOpen-source agent-building library; LangChain 1.0 runs agents on the LangGraph runtime.5
LangfuseOpen-source alternative to LangSmith for tracing and monitoring.1
LangGraphGraph-based framework for stateful, multi-step agents.3
LangSmithLangChain's tracing/eval companion for debugging agent runs.4
LiteLLMOpen-source gateway that gives applications and agents one API in front of many model providers, with key management and routing.9
Meta Muse (personal agent)Meta's personal AI agent for consumers, now paired with a dedicated "Muse Charm" wearable device.14
Meta Waddle World and Style SketchMeta test-market apps: Waddle World, an AI pet companion on iOS and Android in the Philippines, and Style Sketch, which suggests outfits from clothes users own with virtual try-on, in Indonesia, Singapore and Thailand.1
Microsoft Decision-1Microsoft's classification, routing and rubric-evaluation model that returns calibrated probabilities.8
Microsoft Maia 300Microsoft's third-generation in-house accelerator on TSMC N2P: eight HBM4 stacks, 288 GB, 18 TB/s per chip.3
Notion AI agent harnessNotion's rebuilt agent harness (Oct 19 2026): agents get their own computer and browser and can work on a task for up to 24 hours.1
Nvidia DynamoNvidia's AI inference framework.4
Nvidia Open Agent Safety PlatformPlatform unveiled Sep 28 2026 with 100+ industry partners to detect and quarantine rogue AI agents in milliseconds, announced after the recent run of agent-incursion incidents.4
OllamaPopular tool for running open-weight models locally; exposed instances were hijacked by PoeLLM.3
OpenAI Agents SDKOpenAI's own production agent-building toolkit.1
OpenAI CodexOpenAI's coding agent.36
OpenAI Daybreak and Patch the PlanetInitiative in which AI agents uncover critical vulnerabilities and disclosure is coordinated.12
OpenAI Decisions APIOpenAI's fast, cheap decision model became available to all customers on Oct 22 2026 with voice and visual modalities.6
OpenAI DotsAlways-on agents that learn a user's priorities and carry out work autonomously.7
OpenAI misalignment monitoringStaged monitor for unauthorized agent behavior: lightweight checks of internal model signals, automated investigation, then alerts to human responders.2
OpenAI PandoGit-compatible backend built for the volume of changes from engineers and coding agents.1
OpenAI Private IntelligenceEnterprise offering that limits OpenAI staff access to customer data, with zero-data-retention and private-inference options for frontier-model deployments.1
OpenCodeCoding-agent tool whose Console service disclosed on Oct 17 2026 that a "large-scale distillation attack" stole API keys from 1.4% of Console users, now used for unauthorized charges.6
Opera NeonOpera's agentic browser, central to its bet on bespoke browsers for the agentic web.2
Palo Alto Networks ConsoleAI-native platform where users describe a goal in plain language and agents run the analysis and actions.1
Perplexity SPACEPerplexity open-sourced SPACE on Oct 15 2026, a sandbox for containing AI agents.4
REA (Reverse Engineer Anything)Open-source MCP server that connects AI coding agents to reverse-engineering tools such as Ghidra, IDA Pro and Hopper.2
SMART (FAA air traffic tool)AI-assisted tool from Air Space Intelligence that recommends route and trajectory changes to the FAA; rolled out from Sep 2026 and scrutinized after the Nov 13 2026 nationwide ground stop.4
Strands BoxAWS open-source sandbox that confines any agent with OS-level isolation and a temporal policy engine that checks each tool call against the agent's past actions.6
Strawberry (browser agent)Launched AI teammates inside a browser that learn recurring workflows and keep handling browser work without a prompt for each step.1
Stripe Link (agent payments)Stripe's stored-credential network that AI agents use to check out for people; agent-initiated transactions grew 38x in September 2026.4
SurgePixEditable AI-generated visuals plus agent skills that let Claude Code and Cursor call its visual tools.1
Sysdig AI DefenseSysdig added runtime security and governance for AI agents running in production, announced Nov 4 2026.1
TalorysOpen-source personal AI assistant that runs inside the user's own Cloudflare account, installed with one npx command.2
Tencent Yuanbao Search APILaunched Oct 20 2026 with API, MCP and Skill integrations over 157 sources, built on Sogou Search.3
Vercel agent-desktopApache-2.0 Rust CLI and C library from Vercel Labs that lets coding agents drive native macOS, Windows and Linux apps through the OS accessibility tree instead of screenshots.2
Vercel Security ReviewRuns inside Vercel Agent, scans changes on every GitHub push and can commit fixes.2
World ID for Agents (AgentKit)Proof-of-human credentials from Sam Altman's World that let a verified person delegate authority to an AI agent; pitched to banks at Money20/20.3
Zapier and MakeGeneral automation platforms with AI steps bolted on.1
Zest (dining agent)New AI agent that finds restaurants and books reservations over text message.1