Skip to main content

What's New in Version 0.667.0

Released on January 11, 2026

This release supercharges Agentic Tooling with smarter planning, new interactive tools, a refreshed Orchestrator experience, and streamlined naming conventions across the platform! 🚀

🧚 Agentic Tooling Superpowers​

Intelligent Runtime Planning​

Agentic Tooling agents have gained significant new capabilities. Planning for long-running tasks is now determined at runtime - if a single turn is enough to complete the request, the agent will execute immediately without unnecessary planning overhead.

This intelligence extends to tool utilisation: agents now exhaust all possible execution routes via the various tools available, making better use of their full toolkit to accomplish plans effectively.

New Interactive Tools​

Agentic Tooling agents now work more like true collaborators. Two powerful new tools enable real-time human-agent collaboration:

  • ask_user_question - When facing ambiguity or multiple valid paths forward, agents can pause execution and ask clarifying questions before proceeding. Need to choose between approaches? Uncertain about a business requirement? The agent will ask rather than assume, ensuring alignment with your intent.
  • update_plan - Agents dynamically refine their execution strategy as new information emerges. Plans aren't static - they evolve based on intermediate results, tool outputs, and changing requirements.

Granular Tooling & Reasoning Controls​

Agents now have fine-grained control over how they search, analyse, and reason - choosing the right approach for each task:

  • rag - Quick vector search for fast, targeted retrieval. Perfect for customer service and straightforward lookups.
  • deep_rag - Full knowledge hub scan for comprehensive data crunching. Use when you need to analyse across your entire knowledge base.
  • internet_search - Quick web search for timely information.
  • deep_internet_search - Exhaustive web research with multiple queries and source synthesis.
  • analyze - Dedicated file analysis for processing documents, spreadsheets, and uploaded content.
  • reasoning - Structured logical reasoning for problem-solving.
  • deep_reasoning - Extended reasoning chains for complex, multi-step analysis.

This granular control means agents intelligently match their approach to the task - a quick FAQ lookup doesn't trigger a full knowledge base scan, while a comprehensive research request gets the deep analysis it deserves.

Note: The greeting tool has been decommissioned. Agents now handle greetings naturally as part of their conversational flow.

Real-Time Feedback & Intervention​

You're always in control. While an agent executes a multi-step plan, you can:

  • Provide feedback mid-execution - Send guidance while the agent works. Your message is queued and seamlessly incorporated into the agent's reasoning - with full visibility into how your input shaped the next steps
  • Interrupt and correct - Spot something wrong? Step in immediately and the agent adjusts course while preserving execution state
  • Collaborate on decisions - The agent surfaces decision points, you provide direction, execution continues seamlessly

This is what separates true agentic AI from static workflows or chatbots with retrieval. Agents reason, adapt, and collaborate - autonomous when confident, inquisitive when uncertain, and always responsive to human guidance.

Backwards Compatibility: To maintain compatibility with existing workflows or automation processes where you don't want agents asking questions, simply untick "Allow ask user question" under Advanced Settings.

💡 Orchestrator Agents Redesigned​

Brand New Visual Experience​

Orchestrator agents now feature a completely redesigned interface inside ToothFairyAI, giving you better visibility and control over long-running plans.

What's New:

  • Step-by-Step Visibility - See each step of your plan clearly laid out
  • Decision Points - Easily identify where the agent made key decisions
  • Agent Assignments - See which agents are assigned to which tasks at a glance
  • Progress Tracking - Monitor execution progress in real-time

More features are planned around Orchestrator agents - stay tuned!

💬 In-Chat Agent Controls​

Toggle Capabilities On-The-Fly​

Admin and AI Engineer users can now enable or disable agent functionalities directly from the chat interface - no need to navigate away to settings.

How It Works:

  • Tools Icon - Click to view all tools assigned to the agent and toggle them on or off in real-time
  • Knowledge Hub Icon - Click to see all topics assigned to the agent and manage access instantly

Located at the top-right of the chat screen, these controls give you immediate visibility into what an agent can do - and the power to adjust it mid-conversation.

✨ Naming Convention Updates​

Functions → Tools​

We've renamed Functions to Tools across the platform. This naming convention adjustment allows us to better scale features around tools moving forward.

Agent Tools → Agent Capabilities​

Similarly, Agent Tools has been renamed to Agent Capabilities. No behaviour changes - just clearer terminology.

Expanded Tool Creation​

The Tools creation section now offers a cleaner dropdown with all available tool types:

  • API Endpoint - Connect to external APIs
  • DB Procedure - Execute database procedures
  • Python Code - Run custom Python scripts
  • Remote MCP - Connect to MCP servers
  • HTML iFrame - Embed web content
  • Chat Script - Create scripted interactions

🛠️ Improvements​

  • HCL Syntax Highlighting - Added support for HashiCorp Configuration Language (HCL) syntax highlighting in code blocks. Terraform and OpenTofu users rejoice!
  • Charts & UML Rendering - Improved handling of charts and UML diagrams during Agentic Tooling task executions. Visualisations now render cleanly inline, making complex outputs easier to read and act on.

📋 Summary​

This update includes:

  • Intelligent Planning - Runtime decisions for single-turn vs multi-step execution
  • Interactive Tools - New ask_user_question and update_plan tools for true human-agent collaboration
  • Granular Controls - Fine-tuned retrieval (rag/deep_rag) and reasoning (reasoning/deep_reasoning) for optimal task matching
  • Real-Time Feedback - Provide guidance, interrupt, and correct agents mid-execution without losing state
  • Orchestrator Redesign - Fresh UI showing decision points, agent assignments, and step-by-step progress
  • In-Chat Controls - Admins and AI Engineers can toggle tools and knowledge topics directly from chat
  • Naming Updates - Functions → Tools, Agent Tools → Agent Capabilities for better platform scaling

Thank you for using ToothFairy! We're constantly working to improve your experience. If you have any feedback or encounter any issues, please don't hesitate to reach out to our support team at support@toothfairyai.com.

Keep building magic with AI! 🧚‍♀️✨

v0.667.0