AI.info
AI Tools Updates — Page 92
What changed in the AI tools this week — dated, from the makers’ own changelogs.
En anglais
- ThoughtSpot Analytics — Improved liveboard performance and filter response times for complex models when rendering Liveboards and processing interactive filter selections.
Improved liveboard performance and filter response times for complex models.
- ThoughtSpot Analytics — When you publish an object from the primary org to one or more orgs, ThoughtSpot no longer sends notification emails to the admins of the receiving orgs.
ThoughtSpot no longer sends notification emails to admins of receiving orgs when you publish an object.
- LangChain — Creating a public trace or thread share link now requires permission to read the runs, including their inputs and outputs.
- LangChain — The LangSmith MCP adds a query_runs tool that accepts the same parameters as the SDK’s client.runs.query() , including typed selects for costs, token breakdowns, and full inputs and outputs, with curs
The LangSmith MCP adds query_runs with typed selects and cursor pagination. It returns token counts and USD costs by default.
- LangChain — Costs now resolve for every OpenAI model on the pricing page, including those listed only under “All models”, such as GPT-5.6 Luna and Terra. The daily price sync previously read only the models the p
Costs now resolve for every OpenAI model on the pricing page, including those listed only under “All models”.
- LangChain — When a trace or thread filter fails, LangSmith shows the error below the filter and keeps the selected run details open, so you can edit or clear the filter without leaving the page.
When a trace or thread filter fails, LangSmith shows the error below the filter and keeps the selected run details open.
- LangChain — The Messages view now shows the user message on OpenAI Responses API calls that send it as a plain string, which is how calls chained with previous_response_id or a conversation send each new turn.
The Messages view now shows the user message on OpenAI Responses API calls that send it as a plain string.
- LangChain — The Messages view requests smaller pages, so long threads no longer fail with a response size error.
- LangChain — The LangSmith Chat model picker now offers Claude Sonnet 5.5. Sonnet 5 stays the default.
- LangChain — Costs now resolve for the open models Vertex AI hosts through Model-as-a-Service: DeepSeek, Qwen, GLM, Kimi, MiniMax, and GPT-OSS. Learn more .
Costs now resolve for the open models Vertex AI hosts through Model-as-a-Service: DeepSeek, Qwen, GLM, Kimi, MiniMax, and GPT-OSS.
- Figma AI — Connect multiple GitHub organizations to Figma
You can connect multiple GitHub organizations to a single Figma plan.
- ElevenLabs — ElevenAgents
Update agent and Create agent draft now compile and validate structured procedures automatically; merge proposals are available in alpha.
- CivitAI — Challenge Prizes: Claim Them and Pick Your Buzz
Challenge winners now claim their prize instead of having it dropped straight into their account. Crucible winners work the same way.
- Riverside.fm — Get camera ready before you record
Look your best before you record, with live framing and lighting tips and an on-screen ring light.
- Riverside.fm — Test your in-person setup on Mac
Run a quick test before your in-person session to make sure your Mac can handle your camera setup.
- Letta — Release v0.34.4
Automatic reply relay was removed from core channels.
- Letta — Release v0.34.3
Added a read-only file policy, nested subagent launches, and automatic inheritance of parent MCP servers.
- Tabnine — v6.6.8
Fixed inaccurate Gemini usage and cost data; fixed a compatibility issue between the OpenCode IDE agent and the required Tabnine CLI version.
- Ask Sage — v3.9.8
Retired discontinued Google image models (use Nano Banana 2 / Pro)
- LangFlow — v1.13.0.dev32
Langflow nightly v1.13.0.dev32
- LiteLLM — v1.104.0
Claude Opus 5.5 and GPT-6, master key enforcement, stdio MCP off by default
- Taskade — v6.270.0: TSK-Thinking Returns & Protect Files From AI Edits
TSK-Thinking returns to the model menu, TSK-Reasoning thinks longer, and files can be protected from AI edits.
- Taskade — v6.270.1: Taskade EVE Build Hotfix
- Factory Droids — CLI v0.233.0, Desktop v0.190.0: Resume with a message, Script output trimming, plus model window, custom model, and app fixes
Resume commands can send a message in the session, and oversized Script results preserve their start and end.
- LiteLLM — v1.104.0: Claude Opus 5.5, GPT-6, Master Key Enforcement & Team Routing Controls
Claude Opus 5.5 and GPT-6, master key enforcement, stdio MCP off by default
- Speakeasy SDK Generation — Every identity has a Findings page listing that person's risk findings
Each identity now has a Security Findings page listing and filtering that person's risk findings.
- Speakeasy SDK Generation — Install an MCP server into ChatGPT Desktop, and open an install link shared from anywhere
MCP server install pages now set up ChatGPT Desktop and open when shared from documents, wikis, or chats.
- Devin — September 30, 2026
Code blocks and terminals now use Devin Mono, and the live shell preview is read-only.
- Devin Desktop — September 30, 2026
The app now uses Devin Mono in code blocks and terminals, and the live shell preview is read-only.
- Voiceflow — Secrets in handoff steps
A Live agent handoff step can now take its value from a project secret, not just a variable.
- Gram — Install an MCP server into ChatGPT Desktop, and open an install link shared from anywhere
- Gram — Every identity has a Findings page listing that person's risk findings
Each identity has a Findings page listing that person's risk findings, with category and rule filters.
- LangFlow — v1.13.0.dev31
Langflow nightly v1.13.0.dev31
- Relevance AI Workforce — MCP server access in Invent
Invent can call tools on your project’s MCP servers, and you can approve a tool for a single chat for 24 hours.
- Relevance AI Workforce — Decision models added to Workforce conditions and Tools
Jev 1.13 is now available in Workforce condition nodes and in a new Decide tool step.
- Relevance AI Workforce — Dark mode now available
Relevance AI now has a dark theme across the app, including Invent.
- RunPod Serverless — AstaBrief 8B on Serverless
AstaBrief 8B from Allen AI is available on Runpod Serverless with a validated vLLM recipe.
- Abnormal Inbound Email Security — Weekly Detection Enhancements - Week of October 2
Enhanced detection of lateral phishing from compromised senders, strengthening protection for internal-to-internal messages
- Botpress — October 2, 2026
Custom ticket fields can be hidden in ticket details; CSAT surveys are localized; Desk bot integrations can be installed with manual configuration.
- Snowflake Cortex Analyst — Snowflake Native Apps: Observability for Cortex Agents
Observability for Cortex Agents.