GaosCode/PlanWeave

PlanWeave is a file-backed loop engineering system for long-running coding agents. It turns fuzzy plans into claimable tasks, routes them through implementation and review agents, records every run, and keeps the loop recoverable.

What it solves

PlanWeave addresses the fragility of using simple chat interfaces for long-running software engineering tasks. It prevents the "fuzzy" nature of chat-based planning from leading to lost context or unrecoverable errors by transforming goals into a structured, file-backed task graph that can be tracked, reviewed, and recovered.

How it works

PlanWeave converts a project goal into a graph of nodes (files) and blocks (documents). Each block represents a specific task that can be claimed by a focused AI agent. The system routes these blocks through a workflow of implementation and review agents, recording every run as a durable artifact in the local workspace. It supports multiple agent executors (such as Claude Code, Codex, and Grok) and provides a coordinator agent to manage the dispatching of work and recovery of stale states.

Who it’s for

It is designed for developers and engineering teams working on complex coding projects that require parallel implementation, staged checks, and a rigorous review-feedback loop that exceeds the capabilities of a standard LLM chat session.

Highlights

  • File-backed Graph Model: Plans are stored as files and documents, making the project model durable and inspectable rather than ephemeral chat history.
  • Multi-Agent Routing: Different blocks can be routed to different AI agents (e.g., Codex for one task, Claude Code for another) or local scripts for deterministic checks.
  • MCP Integration: Includes a Model Context Protocol (MCP) server allowing ChatGPT to author and manage plans directly.
  • Visual and CLI Interfaces: Offers both an Electron-based desktop canvas for visual planning and a CLI for terminal-driven workflows.
  • Auto-Run Workflow: Capable of automatically claiming blocks, running agents, collecting reports, and handling review feedback loops.
  • Collaborative Execution: Supports shared workspaces and "Agent Hosts" to run agents on remote devices.

Related

  • Project
    PlanExeOrg/PlanExe

    An open-source planning tool that turns complex project briefs into structured, editable planning baselines and interactive reports using AI.

  • Project
    iwe-org/iwe

    A memory system that turns local Markdown files into a queryable knowledge graph for humans and AI agents, providing structured retrieval via CLI and MCP.

  • Project
    OthmanAdi/planning-with-files

    A persistent file-based planning system for AI agents that stores goals and progress in Markdown files to prevent memory loss during context resets.

  • Project
    wandb/weave

    A toolkit for developing Generative AI applications that provides tracing, logging, and evaluation tools to bring rigor to the LLM development process.

  • Project
    akitaonrails/ai-memory

    A long-term memory system for AI coding agents that creates a shared, persistent Markdown wiki from session observations to enable seamless context handoffs between different agents.