ShenSeanChen/waku-agent

Waku Waku! Waku Agent is a local-first AI agent harness you actually own, including loop, memory, eval, all in code built to stay legible as it grows.

What it solves

Waku is a local-first personal AI assistant designed to be transparent and owner-controlled. Unlike commercial AI products, it provides a codebase that users can fully read, modify, and own, avoiding the complexity of heavy frameworks while implementing the core pillars of a professional agent: harness, loop, memory, and evaluation.

How it works

The project is built around four main components:

  • Harness: Manages gateways for interaction (terminal, dashboard, voice, Telegram, Discord, WhatsApp) and the tools surrounding the loop.
  • Loop: A concise Python implementation (approximately 95 lines) that handles the reason-act-repeat cycle.
  • Memory: A local SQLite-based system supporting semantic, episodic, and procedural (skills) memory. It uses a "gate" to decide if information should be remembered and a consolidation process to determine what to keep.
  • Eval/LLM-Ops: Includes deterministic tests and LLM-as-judge evaluations, a release gate, and tracing for every turn.

It supports multiple LLM providers (Anthropic, OpenAI, Gemini, DeepSeek, etc.) via a simple adapter and can optionally connect to a hosted shared memory service (Waku Memory) to sync data across different agents.

Who it’s for

Developers and AI enthusiasts who want a lightweight, readable codebase to build or study how AI agents work without being buried in framework abstraction.

Highlights

  • Local-first: Memory is stored in a single, accessible SQLite file.
  • Local Dashboard: A browser-based cockpit to visualize message flow and agent thinking.
  • Transparent Loop: The core agent logic is written in plain Python for easy stepping-through.
  • Multi-provider support: Works with various hosted models via a single key.
  • Built-in Evals: Combines deterministic and LLM-based testing to ensure reliability.

Related

  • Project
    zhayujie/CowAgent

    CowAgent is an open‑source, self‑hosted AI assistant that plans tasks, runs tools, keeps a personal knowledge base, and can operate as a team of cooperating agents. It supports many LLM providers, multimodal input, dozens of chat channels (Telegram, Slack, WeChat, etc.), a plug‑in skill system, and a three‑tier memory with nightly “Deep Dream” distillation. Install via a one‑line script or Docker, manage everything through a web console, and extend the system through the public Skill Hub or custom tool integrations.

  • Project
    Pinvou/pinvou-agent

    Pinvou Agent is a cross‑platform desktop AI workspace built with Tauri 2, React and a Rust‑based agent engine (CodeWhale). It lets users attach files, run local or cloud LLMs, and use extensible tools to turn chat prompts into real deliverables—documents, designs, or code—while keeping data private when desired.

  • Dispatch
    Waku: A Native GPU-Accelerated Interface for Coding Agents

    Waku is a native Rust application built with GPUI that provides a unified, local-first interface for managing multiple coding agent CLIs through a single timeline and integrated git-based versioning.

  • Project
    hhyqhh/inno-agent

    Inno Agent is a personal learning assistant that uses a three-layer memory system and a proactive scheduler to provide personalized, evidence-driven educational support.

  • Project
    Dicklesworthstone/pi_agent_rust

    A high-performance, native Rust CLI agent for AI-assisted coding that prioritizes low resource usage, secure tool execution, and predictable terminal interaction.