LMQL - Query language and runtime that adds variables, constraints, and control flow to prompting for reproducible LLM programs

Guidance - Template-driven prompting library that interleaves generation with control logic to produce reliable structured outputs

Claude-Mem - Run a local memory service that captures tool usage and summaries so Claude can recall project context across sessions.

Chatterbox - Drop in open-source TTS models to synthesize natural speech for realtime agents and apps.

Goose - Run an on-machine, extensible coding agent that can create projects, execute code, and use tools via MCP.

Open Notebook - Self-host a privacy-first alternative to Notebook LM for organizing sources and grounding AI answers.

Letta - Build stateful agents with long-term memory and a clean API that persists behavior across sessions and models.