Skip to main content

What's New in Version 0.700.0

Released on July 19, 2026

This release introduces Triggers — event-driven automation that runs an agent the moment something changes — alongside a major upgrade to TF Code's self-improving skills, and a heads-up that Kimi K3 is coming really soon! 🚀

🎯 Triggers​

Event-Driven Automation​

Triggers automatically run an agent when something happens — no schedule required. Where a Job runs an agent on a fixed clock, a Trigger runs an agent in reaction to an event, letting your workspace respond the instant something changes.

What's New:

  • Native Events - React to ToothFairyAI's own platform events, including document.created, document.updated, agent_completed, agent_failed, planner_pending_approval, planner_stopped, and more
  • External Providers - Trigger from 30+ third-party apps, including Google Drive, Slack, GitHub, Salesforce, Stripe, Notion, Jira, HubSpot, Linear, Gmail, Google Calendar, Microsoft Teams, and a Generic REST option for any custom API
  • Data Filters - Narrow Native triggers with field / operator / value rules (equals, not equals, includes, greater than, less than, exists) so the agent only fires on events it truly needs to act on
  • Forced Prompts - Skip the prompt dropdown and pass free-text instructions (up to 1024 characters) directly to the agent
  • Polling Acquisition - External triggers poll the provider on a configurable interval (1, 2, 5, 10, or 30 minutes) using a stored Authorisation — webhooks are on the roadmap
  • ARMED / PAUSED Status - Arm a trigger to fire on matching events, or pause it to keep the configuration intact while you re-tune

How It Works:

  • Pick a source - Choose a Native platform event or an external provider event
  • Bind an agent & prompt - Select a text-capable agent and one of its assigned prompts (or a forced prompt)
  • Filter & arm - Add data filters for Native events, set the poll interval for external ones, and toggle the trigger active

📖 Documentation: Triggers

Use Cases:

  • Support triage - zendesk.ticket.created → an agent drafts a first-pass response
  • Sales follow-up - hubspot.deal.created / salesforce.opportunity.updated → an agent drafts a tailored follow-up email
  • Code review assist - github.pr.opened → an agent reads the diff and posts a review checklist
  • Payment alerts - stripe.charge.dispute.created → an agent summarises the dispute and drafts a customer response
  • Document lifecycle - document.created → re-index or summarise a document the moment it lands

🖥️ TF Code​

Self-Improving Skills, Refined​

TF Code's self-improving skills have been expanded with an explicit insight checklist, a refinement flow, and new configuration options that give you far more control over how agents capture and validate reusable knowledge.

What's New:

  • Insight Checklist - Before declaring a task done, the agent checks for a user correction, a repeated pattern or gotcha, or a stale/misleading skill — and only refines when one clearly applies
  • Refinement Flow - skill_manage search finds an existing skill to update (near-duplicates are blocked and redirected), then create or update stamps provenance (source, reason, updated_at, scope)
  • Validation Gate - New validate field runs a judge pass before a self-improved skill is committed — using your reviewer model if configured, otherwise a rule-based gate
  • Version History - Every update snapshots the previous version to a configurable history_dir (outside your working repo); roll back with skill_manage history and skill_manage restore
  • Reserved Embeddings - The default matcher is a local, dependency-free hashed n-gram matcher that works offline, with WASM embeddings planned as an opt-in

New Config Fields:

  • validate - Run a judge pass before committing a self-improved skill (defaults to false)
  • history_dir - Directory for skill version snapshots (override with the TFCODE_SKILL_HISTORY_DIR env var)
  • embeddings - Reserved embedding backend config (default: local n-gram matcher)

📖 Documentation: Self-improving skills

🤖 Models​

Kimi K3 is Coming Really Soon​

We're putting the finishing touches on Kimi K3, the next-generation model from Moonshot AI. Kimi K3 brings strong agentic reasoning and tool-use capabilities to ToothFairyAI, and will be available across chat, agents, and the Agents Arena very soon. Watch this space — we'll have full pricing and availability details the moment it lands.

📋 Summary​

This update includes:

  • Triggers - Event-driven automation that runs an agent on Native platform events or 30+ external providers, with data filters, forced prompts, and ARMED/PAUSED status
  • Self-Improving Skills (TF Code) - Insight checklist, refinement flow, validation gate, version history, and reserved embeddings config
  • Kimi K3 - Moonshot AI's next-gen model is coming really soon

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! 🧚‍♀️✨

v0.700.0