New AI and agent tool releases this week
Last 7 days (Oct 21, 2026 to Oct 27, 2026, as of Oct 27, 2026): 14 tool releases (Goodcall, Jobber Receptionist, LangGraph, Claude Code, Codex CLI, CrewAI, Gemini Notebook, goose, Luma, Replit, Google ADK, opencode, Claude and Mastra) and 4 model releases (GPT-6.1 Terra, Kimi K3.1 Pro, Kimi K3.1 Pro Highspeed and Nemotron 3.5 Bolt E4B (4GB)). 8 tools are marked Tested (a small smoke test with its scope stated); everything else is Reported from official sources and we say so on each page.
Tool releases from Oct 21, 2026 to Oct 27, 2026
| Tool | Release | Date | Evidence | Verdict |
|---|---|---|---|---|
| Goodcall | new-tool | 2026-10-25 | Reported (not run by us) | A sensible first AI receptionist for a one-person shop; the per-customer pricing is simple, but test call quality yourself. |
| Jobber Receptionist | new-tool | 2026-10-25 | Reported (not run by us) | Best if you already use or plan to use Jobber for scheduling and invoicing; the receptionist works with your Jobber data. |
| LangGraph | 1.2.17 | 2026-10-24 | Tested (smoke test, 2026-10-25, local 3B model (Qwen2.5-3B)) | A dependable foundation for stateful agents; this release is a routine fix, so only update if the RemoteGraph or recursion fixes affect you. |
| Claude Code | v2.1.310 | 2026-10-23 | Tested (smoke test, 2026-10-25, local Qwen2.5-3B-Instruct, not vendor models) | Update now if you use the sandbox or run Claude Code on a cloud provider; skip if you wanted flashy new features, this is a hardening release. |
| Codex CLI | 0.164.0 | 2026-10-23 | Tested (smoke test, 2026-10-25, local Qwen2.5-3B-Instruct, not vendor models) | A quality-of-life release: worth updating if you re-type the same prompts or switch models a lot, not a reason to switch tools. |
| CrewAI | 1.15.30 | 2026-10-23 | Tested (smoke test, 2026-10-25, local Qwen2.5-3B-Instruct, not vendor models) | Reasonable if you like the crew-of-agents mental model; the memory and tool-loop fixes in this release are worth taking. |
| Gemini Notebook | update | 2026-10-23 | Reported (not run by us) | A good fit for turning a pile of documents into drafts and decks you can check against sources; availability of new features is rolling out. |
| goose | v1.56.0 | 2026-10-23 | Tested (smoke test, 2026-10-25, local Qwen2.5-3B-Instruct, not vendor models) | Pick goose if you want an open-source agent with real subagent controls; v1.56.0 is a solid, subagent-heavy update. |
| Luma | update | 2026-10-23 | Reported (not run by us) | A good pick for creators who want several video models in one workspace; Ray 3.2 launched on June 9, 2026, so check which models are live on your plan. |
| Replit | update | 2026-10-23 | Reported (not run by us) | Strong pick if you want build, hosting and monitoring in one place; read the credit and overage rules first. |
| Google ADK | v2.13.0 | 2026-10-22 | Tested (smoke test, 2026-10-25, local Qwen2.5-3B-Instruct, not vendor models) | Update with care: a big, useful release for Gemini users, but it includes a telemetry default change. |
| opencode | v1.19.4 | 2026-10-22 | Tested (smoke test, 2026-10-25, local Qwen2.5-3B-Instruct, not vendor models) | Update when convenient for the runtime security fixes; pick opencode if you want an open-source agent that is not tied to one model vendor. |
| Claude | update | 2026-10-21 | Reported (not run by us) | A strong writing partner; try Claude Duet only if you are on Pro or Max and comfortable with a research-preview product. |
| Mastra | @mastra/core@1.79.0 | 2026-10-21 | Tested (smoke test, 2026-10-25, local 3B model (Qwen2.5-3B)) | A strong TypeScript option if you are building agents in a Node stack; read the breaking-change note on sessions before upgrading. |
Model releases from Oct 21, 2026 to Oct 27, 2026
| Model | Vendor | Release date | Evidence |
|---|---|---|---|
| GPT-6.1 Terra | OpenAI | 2026-10-26 | Reported |
| Kimi K3.1 Pro | Moonshot AI | 2026-10-23 | Reported |
| Kimi K3.1 Pro Highspeed | Moonshot AI | 2026-10-23 | Reported |
| Nemotron 3.5 Bolt E4B (4GB) | NVIDIA | 2026-10-23 | Reported |
Sourced facts for each model are on its page. All months and weeks: model release pages.
New and re-verified guides
- How to add an MCP server to Claude Code (Reported (not run by us): from the Claude Code docs; published 2026-10-26)
- Best model for OpenCode, Cline and Claude Code: options by budget (Reported (not run by us): prices and setup from vendor docs; our tool tests do not measure model quality; published 2026-10-26)
- How to build an MCP server with FastMCP (Tested (smoke test, 2026-10-25, no LLM involved), stdio only; published 2026-10-26)
- ChatGPT Work vs ChatGPT Chat vs Codex: what each is for (Reported (not run by us): from OpenAI's documentation; we have not used ChatGPT Work; published 2026-10-26)
- Is Claude Opus 5.5 free? Which plans include it (Reported (not run by us): from Anthropic's pricing page; we did not test access with an account; published 2026-10-26)
- Figma MCP server: how to enable it (Reported (not run by us): from Figma's Help Center and developer docs; published 2026-10-26)
- Gemini 4 Argon API pricing: $2 and $10 per million tokens until Nov 30, then $4 and $20 (Reported (not run by us): from Google's pricing page; we have not called the API; published 2026-10-26)
- Git MCP server: how to set it up and what it can do (Tested (smoke test, 2026-10-25, no LLM involved); published 2026-10-26)
- GitHub MCP server: how to set it up (Reported (not run by us): from the README and Claude Code docs; published 2026-10-26)
- Which coding agents can use GPT-6.1 Terra? (Reported (not run by us): from vendor docs and changelogs; published 2026-10-26)
- Is Kimi K3.1 Pro open-weight? Status of the Nov 4 promise (Reported (not run by us): from Moonshot's blog and a Hugging Face listing check; published 2026-10-26)
- Playwright MCP server: setup, tools and what we measured (Tested (smoke test, 2026-10-25, no LLM involved); published 2026-10-26)
- How to run Claude Code with a local model (Tested (partial, 2026-10-25): one headless edit, small local model, not Claude models; published 2026-10-26)
- Which Claude model should you use? Haiku 5.5, Sonnet 5.5, Opus 5.5 or Fable 5.1 (Reported (not run by us): Anthropic's docs plus our editorial judgement; published 2026-10-26)
- All guides
Latest news briefs
- OpenAI GPT-6.1 Terra: a $1/$5 computer-use model, and what the launch post leaves out (Reported; first seen 2026-10-26 19:05 UTC)
- Superset is now Paradev: parallel coding-agent app ships desktop v2.0.0 and an $11.5M seed (Reported; first seen 2026-10-26 17:05 UTC)
- Anthropic files S-1 for an IPO on Nasdaq under "ANT", reporting $16.3 billion revenue in six months (Reported; first seen 2026-10-26 15:15 UTC)
- ARTEX: an open-source multi-agent LLM system for autonomous penetration testing in isolated labs (Reported; first seen 2026-10-26 15:35 UTC)
- PaperPilot: an open-source Rust PDF toolkit that runs as an MCP server, CLI and in-browser WASM (Reported; first seen 2026-10-26 12:06 UTC)
- Octopus SI OS: an open-source, self-evolving agentic OS with a Rust kernel, released as a research preview (Reported; first seen 2026-10-26 12:27 UTC)
- SolidCraft: open-source parametric CAD in pure Rust, with an MCP server so AI agents can model parts (Reported; first seen 2026-10-26 12:27 UTC)
- MCP reference servers v2.1.0 fixes eight vulnerabilities in filesystem, everything and fetch (Reported; first seen 2026-10-26 03:48 UTC)
- All news briefs
Goodcall
Goodcall builds an AI receptionist from a plain-English description, with plans from $79 a month billed by unique customer instead of per minute. This is a snapshot of its site as of Oct 25, 2026; we found no dated changelog. Full verdict and setup guide.
- Snapshot (no dated release notes found): Goodcall's agent builder, Gabi, creates and edits your receptionist from plain-English instructions.
- Pricing page: Starter $79/month (100 unique customers, then 79 cents each), Growth $129/month (500 customers, then 26 cents), Scale $299/month (2,000 customers, then 15 cents).
- No per-minute charges; every plan includes a local number, and you can keep your own number by forwarding it.
Jobber Receptionist
Jobber's Receptionist answers calls and texts, books visits, takes work requests and escalates urgent calls; a 'Receptionist Companion' that drafts quotes while you take a call is marked coming soon. This is a snapshot as of Oct 25, 2026. Full verdict and setup guide.
- Snapshot (no dated release notes found): Receptionist answers calls and texts 24/7, books visits using your existing online booking settings and creates work requests with details to quote.
- Call escalation: you choose urgent keywords such as burst pipe or no heat and it puts the caller through or texts you an alert.
- Conversation log with summary, transcript and recording; you can watch text conversations live and take over.
LangGraph 1.2.17
LangGraph 1.2.17 is a small fix release: it applies an explicit recursion limit equal to the default and keeps the caller's namespace in RemoteGraph streaming, plus dependency bumps. Full verdict and setup guide.
- 1.2.17 (Oct 24): Apply an explicit recursion_limit equal to the default.
- 1.2.17: Keep the caller's namespace in RemoteGraph.stream(subgraphs=True).
- 1.2.17: Remove precautionary third-party upper bounds on dependencies and bump certifi.
Claude Code v2.1.310
Claude Code 2.1.309 and 2.1.310 are heavy on sandbox and permission-check fixes, plus the Monitor tool now working on Bedrock, Vertex and Foundry and new Enterprise controls for cloud sessions. Full verdict and setup guide.
- v2.1.310: The Monitor tool and plugin monitors now work on Bedrock, Vertex and Foundry, and when telemetry or nonessential traffic is disabled.
- v2.1.310: Requests no longer fail behind gateways that reject the effort beta value; Claude Code retries without it.
- v2.1.310: Claude Code cloud sessions and Remote Control permissions can be given to Enterprise custom roles.
Codex CLI 0.164.0
Codex CLI 0.164.0 adds numbered /presets slots for reusable prompts, inline arguments for /model and /permissions, image previews in supporting terminals, enterprise MCP sign-in and credential masking. Full verdict and setup guide.
- Save reusable prompts and slash commands in ten numbered /presets slots, with a keyboard menu for quick access.
- Set model, reasoning effort, speed and permission profiles directly with inline /model and /permissions arguments.
- Preview viewed and generated images in the fullscreen transcript on terminals that support Kitty or Sixel graphics.
CrewAI 1.15.30
CrewAI 1.15.29 promotes AgentExecutor out of experimental status and adds LiteLLM-compatible kwargs, Azure DocumentDB vector search and new CLI evaluation features; 1.15.30 followed with a small fix on Oct 24. Full verdict and setup guide.
- 1.15.29 (Oct 23): AgentExecutor is no longer experimental.
- 1.15.29: Support for LiteLLM-compatible kwargs; new Azure DocumentDB vector search integration.
- 1.15.29: crewAI eval can run the crew itself for a verdict from one command and export evaluations as files.
Gemini Notebook
Google added guided creation for Slide Decks and Video Overviews in Gemini Notebook (Oct 19-21) and, on Oct 23, a Connected App so you can manage notebooks from the Gemini app. The help center now calls the product Gemini Notebook. Full verdict and setup guide.
- Oct 19-21: Guided creation for Slide Decks and Video Overviews, with suggested topics and styles from your sources and a Review toggle to check an outline or producer notes before rendering.
- Oct 23: Gemini Notebook is a Connected App in the Gemini app; type @Gemini Notebook to create notebooks, add sources, save notes or retrieve information. On by default except in the EEA.
- Oct 16: Study features: 60-second Short Video Overviews, Interactive Learning Overviews and more quiz formats (short-answer, multi-select, fill-in-the-blank).
goose v1.56.0
goose v1.56.0 focuses on subagents: you can resume failed or cancelled delegated subagents, view subagent sessions in the UI and cap concurrency. It also adds a Windows installer for the desktop app. Full verdict and setup guide.
- v1.56.0 (Oct 23): Resume failed or cancelled delegate subagents.
- v1.56.0: Limit concurrent subagents with GOOSE_MAX_CONCURRENT_SUBAGENTS, and open view-only subagent sessions from the UI.
- v1.56.0: Automatic thinking level operation for improved model reasoning.
Luma
Luma added Kling 4.0 for every customer on Oct 23, next to its own Ray 3.2 model, which offers up to 16 keyframes per clip, 16-bit EXR export and a new API. Full verdict and setup guide.
- Oct 23: Kling 4.0 is now in Luma for every customer, aimed at fast action and big camera moves, with text-to-video and image-to-video in 16:9, 9:16, 1:1 and 21:9.
- Ray 3.2 (announced June 9, 2026): frame-level control with up to 16 keyframes in one clip, up to 20-second 1080p clips, native HDR and 16-bit EXR export.
- Ray 3.2 adds expressive facial performance tracking for up to eight faces and a Reframe tool to change aspect ratio, extend the frame or replace a background.
Replit
Replit's Oct 23 update adds a Monitoring Agent that investigates production issues from logs and metrics and can attempt fixes, plus reusable workspace skills, chat and asset sharing, and Stripe Atlas incorporation help. Full verdict and setup guide.
- Monitoring Agent: investigates production issues using your app's logs and metrics, then attempts and validates fixes. Turn on Monitoring alerts and Auto-fix; publishing fixes automatically needs the experimental Auto-republish setting.
- Agent can create and update reusable workspace skills from your conversations and projects, which you review before saving.
- Share a whole chat or a single asset with selected teammates.
Google ADK v2.13.0
ADK for Python 2.13.0 promotes 12 experimental features to stable, adds a classifier node for fast workflow routing, live audio with barge-in, and MongoDB session and memory services. Full verdict and setup guide.
- 2.13.0 (Oct 22): 12 mature experimental features are now stable.
- New TypesafeClassifier and ClassifierNode route workflows with calibrated probabilities in one forward pass instead of an LLM call.
- Live audio interactions with barge-in alongside Gemini speech-to-text and text-to-speech.
opencode v1.19.4
opencode v1.19.4 is a small maintenance release: it upgrades the bundled Bun runtime to 1.4.4, which the release notes say includes security fixes. The project itself is very active and widely starred. Full verdict and setup guide.
- v1.19.4 (Oct 22): Upgraded the bundled Bun runtime to 1.4.4, which the notes say includes security fixes for Brotli, TLS, HTTP/1.1, HTTP/2, HTTP/3, WebSocket and fetch, and updates root certificates (NSS 3.131.1).
- The README lists several install routes: an install script, npm, Homebrew, and a desktop app cask.
Claude
Anthropic's Oct 21 release notes announce Claude Duet, a research preview of a proactive agent app for Pro and Max users; earlier in October came Claude Haiku 5.5 and monthly API credits for Max and Team plans. Full verdict and setup guide.
- Oct 21: Claude Duet research preview for Pro and Max: a new app where Claude works as a proactive agent.
- Oct 7: Claude Haiku 5.5, described as the cheapest, fastest and most capable small model Anthropic has released, for high-volume, cost-sensitive tasks.
- Oct 7: Max and Team plans now include monthly API credits for running your own apps and agents on the Claude Platform; claim them by linking a Console organization in billing settings.
Mastra @mastra/core@1.79.0
@mastra/core 1.79.0 adds experimental MCP Events, richer observability (span aggregation and time-to-first-chunk), session-ID based agent controller sessions (a breaking change) and a native filesystem API for sandboxes. Full verdict and setup guide.
- @mastra/core@1.79.0 (Oct 21): Experimental MCP Events let MCP servers publish event streams that clients can poll, stream or receive by signed webhook, with Studio inspection under mastra dev.
- Observability: new aggregateSpans() for per-span analytics such as p95 latency per tool, plus timeToFirstChunkMs fields and richer querySpans filters.
- Agent controller sessions are now addressed by sessionId, allowing many sessions per resource (breaking change).
Earlier weeks
- Week of Oct 12, 2026: 8 releases
- Week of Oct 5, 2026: 1 release
- Week of Sep 7, 2026: 1 release
- Week of Jan 26, 2026: 1 release
- Full archive
FAQ
What AI tools and models were released this week?
For Oct 21, 2026 to Oct 27, 2026: tools: Goodcall, Jobber Receptionist, LangGraph, Claude Code, Codex CLI, CrewAI, Gemini Notebook, goose, Luma, Replit, Google ADK, opencode, Claude and Mastra; models: GPT-6.1 Terra, Kimi K3.1 Pro, Kimi K3.1 Pro Highspeed and Nemotron 3.5 Bolt E4B (4GB). Each links to its own page with sources.
Which of this week's releases did you test?
8 of 14 are marked Tested, meaning a small smoke test with its scope stated on the tool page. The rest are Reported from official sources.
Related
Markdown version for agents: /this-week.md