AI tools
Last scanned 26 September 2026, 18:54 UTC.
Every installable tool this briefing has seen announced, one row per tool.
New since 25 Sep 2026
First recorded after 25 Sep 2026 — what the radar learned since its last scan on an earlier day. Dates here are when the radar first recorded each tool; the catalogue below shows when a source first reported it.
| Seen | Tool | Maker | Install · Runs | Licence · Cost | Reported by |
|---|---|---|---|---|---|
| today | Agent UltraExa released Agent Ultra, a hosted API mode for exhaustive list building and entity enrichment.workflow: data collection — matched: api | Exa | cloud | proprietary | Marktechpost |
| today | perplexity.aiPerplexity.ai announces a new tool for analyzing competitive losses and building sales strategy decks. | Perplexity | — | — | Perplexity |
Overview
- 52tools trackedsince 17 Sep 2026
- 52in the last 30 days44 in the last 7
- 16of those run locallyon your own machine, not a cloud
- 26makerswith at least one tool on record
Every tool
Grouped by the month it was first reported. Rows whose coverage matched the workflow keywords lead, newest first within each band. The link on each name opens the story this briefing read it in.
September 2026 · 52
| Seen | Tool | Maker | Install · Runs | Licence · Cost | Reported by |
|---|---|---|---|---|---|
| today | Agent UltraExa released Agent Ultra, a hosted API mode for exhaustive list building and entity enrichment.workflow: data collection — matched: api | Exa | cloud | proprietary | Marktechpost |
| yesterday | Engine v2LangSmith Engine v2 introduces red teaming and automated fix validation.workflow: orchestration — matched: automated | LangChain | — | — | LangChain |
| 3 days ago | langgraph-clilanggraph-cli version 0.4.32 released with new deployment features.workflow: orchestration — matched: deployment | LangChain | — | — | LangGraph |
| 5 days ago | DramagicByteDance launched Dramagic, an AI platform for producing short dramas from script to video preview.workflow: publishing/video — matched: video | ByteDance | cloud | proprietary · paid | The Decoder |
| 9 days ago | TauGridMicrosoft open-sources TauGrid, a Kubernetes-native stack for GPU AI workloads, deployable via Helm on any Kubernetes 1.30+ cluster.workflow: orchestration — matched: kubernetes, helm | Microsoft | local | open · free | Marktechpost |
| today | perplexity.aiPerplexity.ai announces a new tool for analyzing competitive losses and building sales strategy decks. | Perplexity | — | — | Perplexity |
| 2 days ago | WhiteboardWhiteboard is an open-source desktop app where humans and agents architect software together in a common workspace. | Unknown | binary · local | open | Hacker News AI |
| 2 days ago | Qwen IntelligenceAlibaba released Qwen Intelligence, a modular phone-agent platform, with Honor Magic9 as the first adopter. | Alibaba | local | — | Pandaily |
| 2 days ago | Qwen BookAlibaba Cloud debuted Qwen Book, its first AI agent PC, with a Day One tester program. | Alibaba | local | — | Pandaily |
| 2 days ago | Project SwapAnthropic announces Project Swap, a tool for agent trading. | Anthropic | — | — | Anthropic |
| 2 days ago | Muse CharmMeta announced the Muse Charm, a new AI gadget for consumers. | Meta | store | — | TechCrunch AI |
| 2 days ago | Managed Deep AgentsLangChain launches Managed Deep Agents 0.8 with user-owned credentials, memory, HTTP channels, and web search. | LangChain | — | — | LangChain |
| 2 days ago | Koi.restKoi.rest is a virtual zen garden project that Paul says everyone can use. | Unknown | — | — | Hacker News AI |
| 2 days ago | Horizon StudioMeta announced Horizon Studio, a browser app offering granular controls for AI game creation. | Meta | store · cloud | proprietary | The Verge AI |
| 2 days ago | Horizon CreateMeta announced Horizon Create, a mobile app for creating games with AI prompts on its Horizon platform. | Meta | store · cloud | proprietary | The Verge AI |
| 2 days ago | CoreWeave Mission Control AgentCoreWeave Mission Control Agent is now generally available. | CoreWeave | cloud | — | CoreWeave |
| 3 days ago | hyperbolixhyperbolix is an open-source library for hyperbolic deep learning in JAX, built on Flax NNX. | Unknown | pip · local | open · free | arXiv cs.LG |
| 3 days ago | MetaBrain SD200 UltraInspur Information launched the MetaBrain SD200 Ultra supernode with 128 tightly coupled domestic AI chips. | Inspur Information | — | — | Pandaily |
| 3 days ago | AnyJevNokia open-sourced AnyJev, a Python library that turns open LLMs into calibrated decision models. | Nokia | pip · local | open · free | Marktechpost |
| 4 days ago | vllm-metalvllm-metal brings vLLM's paged, continuously batched serving stack to Apple Silicon for local inference. | vLLM | local | — | vLLM |
| 4 days ago | llm-typesafellm-typesafe 0.1a0 plugin adds support for TypeSafe AI's Jev model. | Simon Willison | — | — | Simon Willison |
| 4 days ago | llmRelease of llm 0.36 library. | Unknown | pip · local | — | Simon Willison |
| 4 days ago | SoL-PiNVIDIA released SoL-Pi, a set of harness mechanisms for the open-source Pi coding agent. | Nvidia | — | — | Marktechpost |
| 4 days ago | Rabbit OS3Rabbit OS3 launches as a cross-platform AI agent app. | Rabbit | hybrid | — | Wired AI+1 |
| 4 days ago | Paper2AgentPaper2Agent is an open-source framework turning papers into interactive AI agents. | Unknown | — | — | IEEE Spectrum AI |
| 4 days ago | MiMo-V2.6Xiaomi open-sources its MiMo-V2.6 series for reinforcement learning. | Xiaomi | — | — | DigiTimes |
| 4 days ago | GitHub Copilot for JetBrainsGitHub Copilot for JetBrains 1.18.0 adds AI-assisted tool approvals and shared skills. | GitHub | extension · local | — | GitHub Copilot |
| 4 days ago | GitHub Copilot CLIGitHub Copilot CLI now supports faster C++ code intelligence with whole codebase indexing. | GitHub | — | — | GitHub Copilot |
| 4 days ago | DropDrop is a rootless Linux sandbox with gVisor support. | Unknown | pip · local | — | Hacker News AI |
| 4 days ago | Claude CodeClaude Code v2.1.280 adds new features including Claude Opus 5.5 support. | Anthropic | — | — | Claude Code |
| 4 days ago | AI·rete·RAGAI·rete·RAG is a Rete rule engine that decides and RAG explains why. | Unknown | — | — | Hacker News AI |
| 5 days ago | langgraph-sdklanggraph-sdk version 0.4.5 released with new features. | LangChain | — | — | LangGraph |
| 5 days ago | langgraphLangGraph 1.2.12 released with new features and bug fixes. | LangChain | pip · local | — | LangGraph |
| 5 days ago | ZCodeZhipu open-sourced ZCode, an AI coding workbench with desktop, browser, and terminal-agent components. | Z.ai | — | open | TechNode+1 |
| 5 days ago | Strands HarnessStrands Harness is an open‑source agent harness from AWS that runs locally or on a cloud provider. | Amazon | hybrid | open | Marktechpost |
| 5 days ago | MiniMax CodeMiniMax open-sourced MiniMax Code, a terminal coding agent available under the MIT license. | MiniMax | — | open | TechNode |
| 5 days ago | ForemergeForemerge 0.5.0 is a local-first MVP for coordinating coding agents. | Unknown | local | — | Hacker News AI |
| 5 days ago | Cloudflare Python WorkersCloudflare Python Workers are now generally available, supporting Python via WebAssembly. | Cloudflare | cloud | — | Simon Willison |
| 6 days ago | llm-keys-uiRelease of llm-keys-ui 0.1 plugin by Simon Willison. | Simon Willison | — | — | Simon Willison |
| 6 days ago | jevalsjevals is a Python library for typed Jev decisions. | Unknown | pip · localpip install jevals | — | Hacker News AI |
| 6 days ago | datasette-explainRelease of datasette-explain 0.2.2 plugin by Simon Willison. | Simon Willison | — | — | Simon Willison |
| 6 days ago | Flet 1.0Flet 1.0 framework released for production web, desktop, and mobile apps in Python. | Unknown | local | — | Marktechpost |
| 6 days ago | EnvHarnessGoogle’s open source EnvHarness lets AI agents train against environments that evolve with them. | — | — | VentureBeat AI | |
| 7 days ago | datasette-auth-githubRelease: datasette-auth-github 1.0 fixes cookie expiration issues for GitHub login on the demo site. | Simon Willison | pip | — | Simon Willison |
| 7 days ago | OpenClawOpenClaw 2026.9.5 ships atomic updates, plugin hot reload, conversation sharing, and expanded GPT Live. | Unknown | — | — | Marktechpost |
| 8 days ago | SGLang v0.5.20SGLang v0.5.20 release includes new models and 713 PRs from 237 contributors. | SGLang | — | — | SGLang |
| 8 days ago | AgentCore runtimeAWS announces the new AgentCore runtime, a capability of Amazon Bedrock AgentCore for production agents. | Amazon | cloud | proprietary | AWS Machine Learning |
| 8 days ago | AI & Economy ATLAS v1.0Google launched AI & Economy ATLAS v1.0, an interactive open-access site to study AI tool usage. | cloud | — | Google AI | |
| 9 days ago | OllamaOllama v0.34.2 added first-run setup options to sign in or continue locally, shared with the desktop app. | Ollama | binary · local | — | Ollama |
| 9 days ago | Muse for MacMeta's Muse is now available on the Mac to work with files and apps. | Meta | local | — | Marktechpost+6 |
| 9 days ago | GO PublicOpenAI and Cooley co-launched GO Public, a product that can draft S-1 filings for SEC submissions. | OpenAI | — | — | The Information |
| 9 days ago | CCGoogle announces new experimental 'CC' AI agent for families. | cloud | — | Ars Technica AI+2 |
How this page is built
Each sweep reads the same candidate pool the briefing is selected from, and a model decides which stories announce a named tool rather than merely discussing one — a distinction a keyword cannot make. What it finds is folded into a cumulative record, keeping the earliest date and counting the outlets. When no model is reachable a stricter pattern stands in, and rows it produced are marked unconfirmed until a later run can check them. Nothing here is a maker announcement: it is what the sources on the source list reported, and a tool no source covered is a tool this page does not know about.
Install method and command, locality, licence and cost are printed only where a source stated them; everywhere else they read unknown rather than guessed. The radar never installs anything — it is a reading list to act on, not a setup script.
Workflow tags come from keywords in the source headline, summary and fetched article excerpt: a tagged row matched those words, and the tag names them. An untagged row matched none, which says nothing about the tool — only that its coverage used none of the keywords.