Skip to content
/
Tier B A4 Markdown scaffold

AiDE

aide · FixingPixels/AiDE · ★ 106 · last commit 2025-01-17

Standardizes AI context management across sessions with a zero-dependency markdown folder convention.

Best whenMulti-expert ChatGPT GPT for initial planning, then any AI reads the resulting .context/ files.
Skip ifIgnoring token limit warnings when loading large .context trees
vs seeds
bmad-method(no personas, no skills).
Primitive shape
No installable primitives
00

Summary

AiDE — Summary

AiDE (AI-assisted Development Environment) is a minimal, dependency-free framework that standardizes AI-assisted development by providing a .context/ folder drop-in with structured templates for project state, tasks, decisions, and session logs — optionally seeded by a companion ChatGPT GPT for initial project planning.

Problem it solves: AI code assistants lack a standardized way to understand project context across sessions; AiDE provides a lightweight convention (.context/current_state.md, roadmap, tasks, decisions, sessions) so any AI assistant can quickly orient itself at the start of each chat.

Distinctive trait: AiDE ships an explicit "Global Rules" block for AI assistants — a paste-ready instruction set for tools that support persistent system-level instructions — bridging the gap between framework convention and AI behavior even in tools without plugin support.

Target audience: Individual developers who want a zero-dependency, copy-paste context framework compatible with any AI assistant, with optional kickoff support from the AiDE Planning GPT.

Production-readiness: Stable but dormant (106 GitHub stars, last pushed January 2025). A companion ChatGPT custom GPT is available for generating initial project planning documents.

Differs from seeds: Nearly identical architecture to agent-os (Archetype 4 — markdown scaffold); AiDE's distinguishing element is the companion ChatGPT Planning GPT that uses a multi-expert "AI Conductor" orchestration pattern to generate the initial planning document, something agent-os does not offer.

01

Overview

AiDE — Overview

Origin

FixingPixels, an independent developer, released AiDE as an MIT-licensed framework in late 2024. It is a single GitHub repo whose payload is a .context/ folder, documentation, and a companion GPT prompt file.

Philosophy

AiDE is built on maximum simplicity: "No dependencies or configuration required. Just drop the .context folder into your project." The framework treats AI context management as a documentation problem — if the AI can read the right files in the right order, it will behave consistently across sessions.

Manifesto-style quotes (verbatim from README)

"⚠ Disclaimer: Some AI-powered code assistants have predefined token limits that could be quickly consumed when reading or writing to .context files. While this framework has proven tremendously helpful, developers should be aware of these limitations."

"Start every session: Begin each chat session by asking your AI assistant to review the files in .context."

"Implementing these rules ensures the assistant operates effectively, maintains consistency, and reduces the effort required to manage project context."

AiDE Planning GPT

AiDE ships AiDE_Planning_Assistant_GPT.md — a full system prompt for a ChatGPT custom GPT that acts as an "AI Conductor" orchestrating specialized expert agents (Product Manager, Tech Lead, UX Designer, QA Specialist, DevOps Engineer) to collaboratively generate a planning_doc.md for new projects. This GPT-side orchestration is the most interesting component — it simulates a multi-agent team in a single ChatGPT session.

Design Choices

  • Drop-in installation: copy .context/ to project root, no setup required
  • Cross-tool compatible: any AI assistant that reads files works
  • Optional Global Rules: a paste-ready instruction block for AI tools with persistent system prompts
  • The context folder is the entire value proposition — there is no runtime, no hooks, no binary
02

Architecture

AiDE — Architecture

Distribution

Standalone GitHub repo. No package manager. Install = copy the .context/ folder.

Install

# Clone, then copy .context/ into your project
cp -r AiDE/.context /path/to/your-project/

Directory Tree

.context/
├── current_state.md       # Overview of project's current technical state
├── roadmap.md             # High-level goals and milestones
├── tasks/
│   ├── TEMPLATE.md        # Task template
│   ├── active/            # Active tasks
│   ├── completed/         # Completed tasks
│   ├── hold/              # Tasks on hold
│   └── planned/           # Planned tasks
├── decisions/
│   └── TEMPLATE.md        # Decision log template
└── sessions/
    └── TEMPLATE.md        # Session log template

# Companion files at repo root
AiDE_Planning_Assistant_GPT.md   # ChatGPT custom GPT prompt
README.md

Required Runtime

None. Plain markdown files.

Target AI Tools

Any (Claude Code, Cursor, ChatGPT, GitHub Copilot, etc.). The optional planning GPT requires ChatGPT/OpenAI API access.

Config Files

None beyond the markdown files themselves. The optional "Global Rules" block is pasted directly into AI tool system prompts.

03

Components

AiDE — Components

Core Files (Templates)

File Purpose
current_state.md Active features, technical state, recent changes — AI updates proactively
roadmap.md High-level goals and milestones
tasks/TEMPLATE.md Task template with status, priority, description
decisions/TEMPLATE.md Architectural decision record template
sessions/TEMPLATE.md Session log template with summary, outcomes, action items

Planning GPT (Companion)

AiDE_Planning_Assistant_GPT.md — Full system prompt for a ChatGPT custom GPT titled "AI Conductor":

  • Acts as orchestrator of specialized expert sub-agents (Product Manager, Tech Lead, UX Designer, QA Specialist, DevOps Engineer)
  • Iterates through 5 steps: User Alignment → Team Creation → Collaborative Problem Solving → Refinement → Conclusive Assistance
  • Produces a planning_doc.md that seeds the .context/ folder

Optional Global Rules

A 9-rule paste-ready block for AI assistant system prompts covering:

  1. Context Awareness (always prioritize .context/)
  2. File Management (proactive updates)
  3. Task Workflow (move to completed/, use templates)
  4. Decision Tracking
  5. Session Management
  6. Roadmap and Current State updates
  7. Consistency and Communication
  8. Error Handling (missing files → generate boilerplate)
  9. Adaptability

No Commands, Hooks, or Scripts

AiDE ships no slash commands, no lifecycle hooks, no Python/bash scripts. It is purely a documentation convention.

05

Prompts

AiDE — Prompts

Excerpt 1: AiDE Planning Assistant GPT System Prompt (verbatim)

File: AiDE_Planning_Assistant_GPT.md Technique: Multi-expert orchestration persona. The GPT acts as a meta-agent that creates specialized sub-agents for different domains, then facilitates collaborative brainstorming among them.

## Objective  
**Act as the AI Conductor**, an orchestrator of specialized AI expert agents. Guide the user to develop a comprehensive **code project planning document** by creating and managing these agents.

## Conductor Process

1. **User Alignment**  
   - Gather context about the project's scope, requirements, and constraints by asking targeted questions.

2. **Team Creation**  
   - Based on the project needs, propose specialized agents (e.g., **Product Manager**, **Tech Lead**, 
     **UX Designer**, **QA Specialist**, **DevOps Engineer**, etc.).  
   - Encourage the user to refine these roles, add new ones, or modify the competencies of any existing agents.

3. **Collaborative Problem Solving**  
   - Facilitate a brainstorming session among the expert agents, focusing on:  
     - **Requirements Gathering & Analysis**  
     - **Architecture & Technology Decisions**  
     - **Project Timeline & Milestones**  
     - **Resource Allocation**  
     - **Risk Assessment & Mitigation Strategies**  
   - Agents should propose and discuss their insights from different angles to shape a planning document.

4. **Refinement Through Feedback**  
   - After presenting the draft of the planning document or major sections, ask the user for feedback.  
   - Based on the feedback, refine existing agents or create new ones to address gaps or newly identified needs.

Excerpt 2: Global Rules Block for AI Assistants (verbatim from README)

Technique: Persistent instruction injection for AI tools supporting system-level rules. Covers 9 behavioral rules establishing the .context/ folder as authoritative project context.

1. **Context Awareness**:
   - Always prioritize the `.context` folder for understanding project details.
   - At the start of every session, load and analyze the latest files from the `.context` directory:
     - `current_state.md`
     - `roadmap.md`
     - All files under `.context/tasks/`, `.context/decisions/`, and `.context/sessions/`.

2. **File Management**:
   - Update the `.context` files proactively when tasks are completed, 
     new milestones are set, or decisions are made.

8. **Error Handling**:
   - If the `.context` folder is incomplete or missing expected files, 
     provide actionable suggestions to generate or recover the necessary files.
   - Proactively generate boilerplate content for missing files, 
     e.g., suggest creating an initial `roadmap.md` for new projects.
09

Uniqueness

AiDE — Uniqueness

Differs from Seeds

AiDE is architecturally identical to agent-os (Archetype 4 — markdown scaffold, zero binaries). Both drop structured markdown folders into a project and rely on the AI to read them. AiDE's two distinguishing elements are: (1) the companion Planning GPT which uses a "AI Conductor" multi-expert pattern to generate initial planning documents — something agent-os does not offer; and (2) the explicit "Global Rules" paste-block designed for AI tools with persistent system-level instructions. AiDE also acknowledges the token consumption problem openly in its README, whereas agent-os does not. Compared to Aegis, AiDE has no YAML operation specs, no self-improvement records, and no task state machine — it is simpler and more portable.

Positioning

AiDE positions itself as "drop-in simplicity" — the lightest possible framework for AI context management. It targets developers who want convention without commitment.

Distinctive Opinion

The companion ChatGPT GPT uses an "AI Conductor" pattern that simulates multi-expert team collaboration within a single chat session — an unusual approach that separates initial project scoping (GPT-side, multi-persona) from ongoing development context (any AI, file-based).

Observable Failure Modes

  1. Token ceiling: The README explicitly warns about context window consumption — a self-acknowledged limitation
  2. No enforcement: No hooks, no scripts, no tests — entirely discipline-dependent
  3. Dormancy: Last updated January 2025; active development appears paused
  4. GPT dependency: Initial planning workflow requires ChatGPT access, which may not match users' tool preferences
04

Workflow

AiDE — Workflow

For New Projects

Step Action Artifact
1 Use AiDE Planning GPT to brainstorm and generate planning doc planning_doc.md at project root
2 Instruct AI to read planning_doc.md and populate .context/ roadmap.md, tasks/planned/*.md
3 Begin development sessions with AI reading .context/ first Updated current_state.md

For Existing Projects

Step Action Artifact
1 Drop .context/ folder into project root Empty .context/ structure
2 Instruct AI to analyze codebase and populate state current_state.md, roadmap.md populated

Session Workflow (Ongoing)

Step Action Artifact
Session Start AI reads current_state.md, roadmap.md, all tasks/decisions/sessions (in memory)
Task Work AI updates task markdown, moves completed tasks tasks/completed/*.md
Decision Made AI documents in decisions/ decisions/DECISION-*.md
Session End AI records session summary sessions/SESSION-*.md

Approval Gates

None explicit. The framework relies on the developer instructing the AI when to update files.

Artifacts Per Phase

Phase Artifact
Planning planning_doc.md (GPT output)
Init roadmap.md, current_state.md
Development tasks/active/*.md
Completion tasks/completed/*.md
Documentation decisions/DECISION-*.md, sessions/SESSION-*.md
06

Memory Context

AiDE — Memory & Context

State Storage

All state is file-based markdown in .context/. No external database, no MCP server.

Files

File Purpose Persistence
current_state.md Technical state snapshot — active features, recent changes Project-scoped
roadmap.md Milestones and goals Project-scoped
tasks/active/ Currently in-progress tasks Project-scoped
tasks/completed/ Done tasks (historical record) Permanent
decisions/ Architectural decisions Permanent
sessions/ Session logs Permanent

Cross-Session Handoff

Yes — the framework's primary value is enabling continuity across sessions. The developer instructs the AI to read .context/ at the start of every session.

Context Compaction

Not handled. No mechanism to summarize or compact growing context files. The README warns this is a concern: "Some AI-powered code assistants have predefined token limits that could be quickly consumed."

Memory Persistence Level

Project-scoped. All memory lives in the git-tracked .context/ folder.

07

Orchestration

AiDE — Orchestration

Multi-Agent

No for the runtime framework. The Planning GPT companion simulates multi-agent behavior (AI Conductor + domain experts) within a single ChatGPT session, but this is a planning-only feature not a runtime harness.

Orchestration Pattern

None (runtime). The Planning GPT uses a sequential multi-persona pattern for initial planning.

Execution Mode

Interactive-loop. The developer instructs the AI manually.

Isolation Mechanism

None.

Multi-Model

No (runtime). Planning GPT requires ChatGPT/OpenAI; the runtime .context/ framework works with any model.

Consensus Mechanism

None.

Prompt Chaining

No.

Crash Recovery

No explicit mechanism.

Context Compaction

No explicit mechanism. The README warns about token consumption.

08

Ui Cli Surface

AiDE — UI / CLI Surface

CLI Binary

None.

Local UI

None.

IDE Integration

None dedicated. Works with any AI assistant that reads files.

ChatGPT Planning GPT

A companion custom GPT at https://chatgpt.com/g/g-67798aff6be881918ac59747a7e6542d-aide-planning-assistant-gpt provides an interactive planning session. This is the only "UI" surface — a ChatGPT chat interface.

Observability

  • sessions/ directory provides a log of all development sessions
  • decisions/ directory provides an audit trail of architectural decisions

No Hooks, No Scripts

AiDE ships no automation whatsoever — everything is driven by human instruction to the AI.

Related frameworks

same archetype · same primary tool · same memory type

BMAD-METHOD ★ 48k
A4 Markdown scaffold

Provides a full agile delivery lifecycle with named expert-persona AI collaborators that elicit the human's best thinking rather…

Agent OS ★ 4.6k
A4 Markdown scaffold

Extracts implicit codebase conventions into token-efficient markdown standards files and injects them selectively into AI agent…

Claude Conductor ★ 367
A4 Markdown scaffold

Gives Claude Code a persistent, cross-linked, auto-analyzed documentation system so it retains codebase context across sessions.

Spec-Driver (Greenfield Spec-Driven Development) ★ 25
A4 Markdown scaffold

Prevents spec rot in AI-assisted development by making implementation changes flow back into evergreen, authoritative specs via…

Anthropic Knowledge Work Plugins ★ 16k
A4 Markdown scaffold

Role-specialized plugin bundles with live MCP connectors that turn Claude into a domain expert for enterprise knowledge workers.

Codex Integration for Claude Code (skill-codex) ★ 1.3k
A4 Markdown scaffold

Single Claude Code skill that handles Codex CLI invocation correctly (stdin blocking, thinking token suppression, session resume)…