v0.15.0

New Features

• Supabase Authentication & User Management: Full Supabase-based authentication with user and team management, enabling secure multi-user access to team resources.
• Standalone Team API: Comprehensive team API with endpoints for agent management, core configuration, metadata, autonomous scheduling, content feeds, and chat — ready for frontend integration.
• JWKS JWT Verification: Team API now supports RS256 JWT verification via JWKS, with automatic key rotation support. Legacy HS256 signing key remains as a fallback.

Improvements

• Improved team API structure with dedicated routers for core and metadata operations.
• Enhanced team membership and permission system.

Full Changelog: v0.14.0...v0.15.0

v0.14.0

New Features

• Team API: Full team-scoped API for agent management, autonomous scheduling, and chat — enabling multi-user teams to manage agents collaboratively via authenticated endpoints.
• Team Content Feed: Teams can now subscribe to agents and receive aggregated activity and post feeds. Teams auto-subscribe to their own agents, and can subscribe to any public agent. Content is distributed in real-time via fan-out-on-write.
• Video Generation Skills: New video generation skill category with support for Grok, Sora (OpenAI), Veo (Google), and MiniMax providers.
• MiniMax LLM Provider: Added MiniMax as a supported LLM provider, including image generation support.

Improvements

• Agent-to-agent calls now propagate attachments from the called agent's response.
• Updated LLM model list with latest provider offerings.

Bug Fixes

• Fixed documentation site domain configuration.

Full Changelog: v0.13.0...v0.14.0

v0.13.0

New Features

• Team Lead Chat: A new "Lead" tab in the frontend lets you chat directly with a team lead agent that manages all your agents. Create, configure, update, and monitor agents through natural conversation.
• Team Channel Integration: Added Telegram team channel integration for team-level communication.
• Image Skills: New image skill category with multi-provider support for image generation and processing.
• Image Attachments in Chat: Chat messages now render image attachments with lightbox preview and download support.
• Custom LLM Providers: Added support for configuring custom LLM providers.
• Team Management: New team management features with invite system and Redis-based caching.

Improvements

• Reorganized skill system for better maintainability — cleaned up unused skills and consolidated categories.
• Simplified LLM model configuration by removing legacy fields.
• Improved agent executor architecture with extracted helpers and reduced complexity.
• Enhanced data integrity with fixes for quota race conditions and credit handling.
• Various stability improvements including cache eviction, timeouts, and resource management.

Bug Fixes

• Fixed issues in the security module.
• Fixed chat cancellation handling to cleanly remove dangling messages.
• Fixed search pricing calculation.

Full Changelog: v0.12.3...v0.13.0

v0.12.3

New Features

• LLM Thinking/Reasoning Display: Agent responses now show the model's thinking process. Thinking content appears as a collapsible block in the chat, giving users insight into how the agent reasons before acting.
• Stop Generation Button: Users can now cancel in-progress streaming responses.

Improvements

• Redesigned chat message layout: thinking and tool call indicators now appear without avatar or background, keeping the conversation view cleaner and more focused on the agent's actual responses.
• Optimized system skill loading for providers with native search capabilities.

Bug Fixes

• Fixed issues with thinking content not displaying for tool call responses.

Full Changelog: v0.12.2...v0.12.3

v0.12.2

New Features

• Read Webpage Skill: Agents can now read and extract content from any webpage using Cloudflare Browser Rendering. The content is automatically cleaned and formatted for better readability. Enabled when the agent's internet search toggle is on and Cloudflare credentials are configured.
• Activity Timeline Cards: The activity timeline now displays rich post cards and link cards with Open Graph metadata previews.

Improvements

• Enhanced input validation for the activity publishing skill
• Improved activity skill prompts for better agent behavior
• Better example agent configuration

Bug Fixes & Maintenance

• Fixed minor issues in the release workflow
• Removed unused debug components
• Updated dependencies

Full Changelog: v0.12.1...v0.12.2

v0.12.1

What's New

• Sub-Agents: Agents can now delegate tasks to other agents with controlled access and configurable timeout
• Skill Category Icons: Skill categories now display icons in the frontend for better visual identification

Improvements

• Skill categories that depend on platform API keys are now properly gated by key availability
• Improved stability for agent delegation calls

Bug Fixes

• Fixed issues in the agent management UI
• Fixed various sub-agent related bugs

Full Changelog: v0.12.0...v0.12.1

v0.12.1-dev1

What's New

• Sub-Agents: Agents can now delegate tasks to other agents with controlled access and configurable timeout
• Skill Category Icons: Skill categories now display icons in the frontend for better visual identification

Improvements

• Skill categories that depend on platform API keys are now properly gated by key availability
• Improved stability for agent delegation calls

Bug Fixes

• Fixed issues in the agent management UI
• Fixed various sub-agent related bugs

Full Changelog: v0.12.0...v0.12.1-dev1

v0.12.0

Release v0.12.0

New Features

• Long-Term Memory: Agents can now maintain persistent long-term memory across conversations. When enabled, agents can store and recall important information using an LLM-powered memory management system that intelligently merges and consolidates memories.

Improvements

• Minor code formatting fixes in tests.

Full Changelog: v0.11.27...v0.12.0

v0.11.27

v0.11.27

Bug Fixes

• Fixed an issue where frontend features (chat, tasks, activities, posts, archive) would fail when navigating via agent slug URLs. After the recent slug URL rewrite feature, some API calls were incorrectly using the slug instead of the agent ID.

Full Changelog: v0.11.26...v0.11.27

v0.11.26

v0.11.26

New Features

• Agent Slug Identifiers: Agents can now be identified by human-readable slugs in addition to their IDs, making agent URLs and references more user-friendly.
• New AI Models: Added Hunter Alpha and Healer Alpha models from OpenRouter, expanding the available model selection.

Improvements

• Improved system prompt generation for better agent behavior.
• Refactored internal engine architecture for clearer naming and better maintainability.
• Fixed a cache invalidation bug where changes to agent data (such as wallet addresses, API keys, or credentials) were not properly triggering agent reloads.
• Removed legacy skill pattern mechanism to simplify the codebase.
• Updated dependencies.

Full Changelog: v0.11.25...v0.11.26