Skip to main content

ToothFairyAI Release Notes

Welcome to the ToothFairyAI release notes section. Here you'll find detailed information about all product updates, new features, bug fixes, and important announcements.

Latest Releases​

Version 0.676.0 - Individual Plan, Agentic Tooling 2.0, Fine-Tuning Expansion & Qwen 3.6​

Released: April 11, 2026

New Individual subscription plan merging Starter and Pro into a flat-rate plan with all features unlocked ($14.99/mo, 1,500 UoI, 15 USD free credits). Agentic Tooling 2.0 (Beta) — an adaptive digital worker with autonomous tool selection, parallel execution, agent delegation, skills, and memory (reasoning models only). Three new fine-tuning methods: GRPO (reward-based RL), KTO (unpaired preference), and CPT (continued pretraining). Self-hosted training pipeline on ToothFairyAI infrastructure with spot instances, VLM fine-tuning, and 35 trainable models. Fine-tuning API with 6 new REST endpoints. Qwen 3.6 Plus with 800K context. New /dispatch async API endpoint. Image, video and 3D generation now available for Operator agents with AT 2.0.

Version 0.675.0 - Release Of Tfcode, Enhanced Import, Antagonistic Benchmarks & Predictions API​

Released: March 29, 2026

Introducing tfcode, your personal AI engineer for ToothFairyAI—natively integrated with MCP and SDK to enhance all development operations across your workspace! Write code using natural language from Python scripts and API integrations to full-stack applications and database queries through simple conversational instructions. Built as a fork of opencode, tfcode is available for anyone with a terminal (Business and Enterprise subscriptions). Enhanced import capabilities for Tools, MCPs and Skills directly from the Tools section for platforms like SharePoint, Salesforce, HubSpot and Twilio. New antagonistic benchmarks for real-world multi-turn scenario testing. Dramatically improved MCP and SDK with better skills management, steering and validation. New /predictions endpoint for direct access to serverless and fine-tuned models. Fixed dashboard loading issues with high-volume UoI consumption data.

Version 0.674.0 - MagicWand API, Fine-Tuning & Twilio Integration​

Released: March 16, 2026

Enhanced Mystica models with improved tool calling capabilities, new MagicWand AI API endpoint for converting conversations into SFT and DPO training data, native fine-tuning capabilities for tool calling and JSON outputs, and native Twilio mobile phone integration allowing businesses to connect any phone number to TF voice agents.

Version 0.673.0 - GraphQL, Enhanced Versioning & Plan Mode Improvements​

Released: March 11, 2026

API Tools now natively support GraphQL queries and mutations with full context-aware features, significantly improved diff UI for reviewing changes across versions, code interpreter includes ToothFairyAI Python SDK by default, Plan Mode now supports multiple choice questions with agent recommendations, Starter plans can now create images, videos and 3D models, Orchestrator agents can be benchmarked like standard agents, and improved voice cloning experience with better visual feedback.

Version 0.672.0 - Voice Languages, Versioning & API Enhancements​

Released: March 3, 2026

Voice agents now speak 15+ languages natively with Australian accents, Agents/Tools/Hooks versioning is production ready for Business/Enterprise, enhanced Sorcerer models for customer service with improved steerability and speed, fixed orchestrator parallel task execution issues, MCP updates for better SDK integration, and improved API tools for binary data conversion including PDF analysis.

Version 0.671.0 - Interleaved Thinking, Voice Channels & Model Management​

Released: February 22, 2026

Enhanced agentic tooling with interleaved thinking for smarter tool calling and MCP handling, streamlined external model management with immediate availability of new OpenAI, Anthropic and Vertex AI models, voice agent native integrations with email, SMS and WhatsApp channels, and fixed knowledge hub document title editing issues.

Version 0.670.0 - Agent Skills, Structured JSON Output & Orchestrator Enhancements​

Released: February 12, 2026

Introduction of Agent Skills for custom agent capabilities via Settings/Tools, Structured JSON Output for schema-compliant responses from Operator, Programmer and Assistant agents, Orchestrator predefined plan schemas for deterministic planning with agentic flexibility, enhanced internet search capabilities with smart filtering, expanded hook environment variables (chat_id, workspace_id, message_id), improved title generation UX, and API expansion for artifact retrieval by chat_id.

Version 0.669.0 - Kimi K2.5 Thinking, Voice Agent Controls & Agentic Upgrades​

Released: February 3, 2026

Introduction of Kimi K2.5 Thinking with record-breaking demand in under 48 hours, strategic model deprecation program for compute optimization, voice agents interface with pre-conversation input/output selection, significant agentic enhancements to Sorcerer and Mystica families, improved code interpreter stability for SQL Server and Oracle DB, and MCP/API expansions for intelligent agent self-optimization.

Version 0.668.0 - Desktop File Access, Advanced Vectorisation & SDK Expansion​

Released: January 17, 2026

Improved planning and tool detection for Agentic Tooling, desktop client can now manipulate files in authorised folders, advanced vectorisation for Business/Enterprise, SDK coverage expanded to 80% of API endpoints (Node.js & Python), enhanced MCP server documentation, and new API for managing secrets for tools and hooks.

Version 0.667.0 - Agentic Tooling Superpowers & Orchestrator Redesign​

Released: January 11, 2026

Intelligent runtime planning, new ask_user_question and update_plan tools for human-agent collaboration, real-time feedback mid-execution, redesigned Orchestrator UI, in-chat controls for toggling tools and knowledge topics, plus naming updates (Functions → Tools, Agent Tools → Agent Capabilities).

Version 0.665.0 - UoI Pricing, MCP Server & Voice Agents GA​

Released: January 4, 2026

Units of Intelligence (UoI) consumption-based pricing, SSE MCP Server for Claude Code and Gemini CLI, Minimax 2.1 & GLM 4.7 models, enhanced agentic tooling with dynamic planning, Voice Agents GA for Business/Enterprise, and Linux desktop client.

Version 0.662.0 - Chat UX, Orchestrators API & Model Updates​

Released: December 21, 2025

Redesigned chat UI with agent switching via # mentions, revamped orchestrators with API access via /planner endpoint, Mystica/Mystica 1.5 intelligence upgrades, GLM 4.6 with fast inference, Basic Auth for agent tools, and reorganised AI API.

Version 0.661.0 - Plan Mode, DeepSeek 3.2 & Llama 4 Finetuning​

Released: December 6, 2025

Plan Mode for Agentic Tooling, DeepSeek 3.2 for all subscriptions, Llama 4 Maverick finetuning for Enterprise, official Python SDK for easy agent integration, and Voice to Voice WebSocket API for third-party platforms.

Version 0.660.0 - Deep Knowledge Search, Analytics & Voice Agents​

Released: December 1, 2025

Deep Knowledge Search for Agentic Tooling, comprehensive Analytics Dashboard with AI-generated insights, enhanced Sorcerer models, refreshed Chat UI, and Voice to Voice agents private beta with Twilio integration.

Version 0.650.0 - MCP Support & Deep Search Upgrade​

Released: November 22, 2025

Remote MCP server support with OAuth2.1/Bearer/API key authentication, massively upgraded Deep Search with agentic loop architecture, high-fidelity image generation, Mystica Vision model, and streamlined Agentic Tooling UX.

Version 0.640.0 - European Launch & Advanced Features​

Released: November 9, 2025

Workspace Auto-Invite for Business & Enterprise, Minimax M2 and Kimi K2 Thinking models, enhanced Dynamic Model Selection, Canvas Selection Awareness, Advanced Prompt Buffering, and European soft launch with complete data sovereignty.

Version 0.632.0 - Enhanced Retrieval & Database Interaction​

Released: October 28, 2025

Re-ranking toggle for improved retrieval accuracy and dynamic database execution capabilities for Operator agents.

Version v0.631.1 - Enhanced Reasoning & Orchestration​

Released: September 30, 2025

Boosted reasoning for Sorcerer and Mystica models, Orchestrator-Operator collaboration with Agentic Tooling, and intelligent topic filtering for knowledge retrieval.

Version 0.628.0 - Agentic Tooling Revolution​

Released: September 22, 2025

Three massive improvements to Agentic Tooling: task scheduling, extended Canvas documents, and intelligent branching task lists.

Version 0.627.3 - Enhanced Collaboration & Security​

Released: September 13, 2025

Canvas for Agentic Tooling, Multi-Factor Authentication, workspace-level MFA enforcement, and direct tool references in chat.

Version 0.626.0 - AI Image Editing & Job Scheduling​

Released: August 30, 2025

AI image editing capabilities, job scheduling for Business/Enterprise, and enhanced NER display.

Version 0.623.0 - Job Scheduling & Workspace Improvements​

Released: August 30, 2025

Job scheduling introduction and improved workspace creation flow.

Version 0.622.1 - New Models & Communication​

Released: August 24, 2025

DeepSeek v3.1 and OpenAI OSS 120B models, enhanced communication capabilities with email/SMS/WhatsApp support.

Release Types​

  • Major Releases (x.0.0) - New features and significant updates
  • Minor Releases (x.y.0) - Feature additions and improvements
  • Patch Releases (x.y.z) - Bug fixes and security updates
  • Hotfixes - Critical security or stability fixes

Stay Updated​

  • Follow us on Linkedin for updates and events