WrongStack/WrongStack

An AI coding agent that reads your code, edits files, runs commands, and reasons through bugs — across a terminal REPL, a full-screen TUI, and a browser UI, while you keep your hand on every permission.

What it solves

WrongStack is an open-source AI coding agent designed to handle complex software development tasks while maintaining a deep, persistent understanding of a specific codebase. It eliminates the need for subscription-based AI tools and avoids the the limitations of "wrapper" tools by providing a standalone engine that manages its own memory, tool execution, and multi-agent coordination.

How it works

WrongStack uses a custom-built kernel (Container, Pipeline, EventBus, RunController) to coordinate a "fleet" of specialist agents under a Director. It features a dedicated memory system called SAGE (using SQLite/FTS5) that anchors project knowledge to files, symbols, and commits. The system includes a Brain for policy decisions, Kanban for task tracking with atomic verification, and Chimera for automated diff reviews. It supports a wide range of LLM providers (Anthropic, OpenAI, Google, Ollama, vLLM) and offers six different user interfaces, ranging from a simple REPL to a full WebUI and a cross-machine command center (HQ).

Who it’s for

It is built for developers who want a powerful, self-hosted AI coding assistant that can operate autonomously on their local machine, manage large codebases with durable memory, and maintain strict permission boundaries over tool execution.

Highlights

  • SAGE Memory: Long-term, code-anchored memory that persists across sessions.
  • Multi-Agent Fleet: A Director agent that can delegate tasks to isolated specialist workers.
  • Standalone Engine: A first-party kernel and tool executor, not a wrapper around other CLIs.
  • Comprehensive Toolbox: 72 built-in tools for editing, testing, linting, and browser/E2E control.
  • Six UI Surfaces: Flexible access via REPL, TUI, WebUI, SimpleUI, Desktop, and HQ.
  • Verifiable Goals: A /goal system that runs until tasks are verifiably complete via Kanban gates.
  • Chimera Review: Automated critique and fixing of file diffs.
  • Local-First: Full support for local LLM endpoints like Ollama and vLLM.

Related

  • Project
    deepagent-ltd/deepagent-code

    DeepAgent Code (DAC) is a desktop‑plus‑terminal workspace that lets you use an LLM as a coding assistant for multi‑step tasks. It offers persistent sessions, inspectable project memory, sub‑agents for independent jobs, a managed browser, and optional instant‑messaging integration, all configurable with any compatible LLM provider. Licensed under AGPL‑3.0.

  • Project
    syncable-dev/memtrace-public

    A structural memory system for AI coding agents that turns codebases into local, bi-temporal knowledge graphs for fast, low-cost architectural queries.

  • Project
    peterfei/ai-agent-team

    A virtual AI development team with specialized roles (PM, Backend, Frontend, etc.) and a persistent memory system for managing complex software projects.

  • Project
    Fergana-Labs/stash

    Stash is an open‑source platform that captures AI‑agent runs, curates the resulting traces into durable knowledge, and lets you package that knowledge as reusable “skills”. It offers a Python SDK, REST API, CLI, and an optional reward‑model loop that turns human‑annotated preferences into prompt‑optimisation signals, enabling continuous improvement of agents without retraining the underlying model.

  • Project
    gi-dellav/zerostack

    zerostack is a lightweight coding agent written in Rust that runs in the terminal, supporting multiple LLM providers. It offers granular permission controls, session management, subagents, and optional persistent memory, while using far less RAM and CPU than JavaScript-based agents.