You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
🐢⚡ Smart API key rotation for Agent Zero. Paste several keys: KAME picks the healthiest one per call, reads every 429 / RESOURCE_EXHAUSTED / quota error, rests each key exactly as long as needed and never lets a rate limit end your run. Gemini, OpenAI, OpenRouter, Anthropic, any provider.
Transcribe YouTube videos, playlists, and entire channels with adaptive rate-limiting to avoid YouTube throttling. Agent Zero plugin with AI-powered summaries, timestamped notes, visual context extraction, and resumable channel jobs.
Automated code pipelines: per-repo worktree bundles plus a shadow repo for project instructions, developer subagent builds, exit-code verification, parallel review by up to 3 agent profiles, a bounded fix loop with feedback to the developer, and push + PR — with terminal-state notifications. Built on stock parallel jobs and the subagents orchestrat
Fully managed Hindsight memory with embedded PostgreSQL, automatic API installation, personal continuity, project-scoped subagent memory, and bounded recall context.
Queryable chat history for every conversation (main chat, subagents, background jobs) in an embedded Postgres database (pg0, no extra containers): full-text and semantic search, context inspection, and a live replica that never loses messages to compaction. Embeddings use the embedding model already configured in model settings. Optional Continuous
Automatic semantic code index for Agent Zero projects. Detects every git repository in the active project folder (whole folder when there are none), indexes incrementally on startup and file changes, and ships the code_search tool: natural-language and symbol search across all repos with an optional repository filter. Embeddings use the embedding m
Manage custom OpenAI-compatible model providers. Add, edit, delete, and test custom LLM endpoints that integrate seamlessly with the model configuration dropdowns.
Portable model transport improvements for Responses, Chat Completions, and Anthropic Messages, with normalized reasoning, native tools, retries, media handling, and provider-safe state replay.
Autonomous heartbeat — wakes the agent after an idle period for silent maintenance work (notes queue, journal, maintenance checks). Replaces the external systemd daemon with an in-process scheduler.