Skip to content
/

Claude Conductor

claude-conductor · superbasicstudio/claude-conductor · ★ 367 · last commit 2026-05-17

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

Best whenThe highest-leverage investment in AI-assisted development is structured project documentation — not workflow automation, slash-commands, or prompt engineering.
Skip ifMonolithic documentation files (Modular > Monolithic), Stale documentation (Maintained > Stale)
Primitive shape
No installable primitives
00

Summary

Claude Conductor — Summary

Claude Conductor is a lightweight, modular documentation framework delivered as an npm CLI tool that scaffolds interconnected Markdown files into any project to improve Claude Code's understanding and navigation of a codebase. It solves the problem of context loss between AI coding sessions by providing a structured, cross-linked set of documentation templates (CONDUCTOR.md, CLAUDE.md, ARCHITECTURE.md, JOURNAL.md, TASKS.md, and up to 14 total files) that Claude Code reads automatically. What makes it distinct is its combination of a zero-network-request offline CLI with an opinionated journaling and error-tracking protocol — specifically the auto-archiving JOURNAL.md and the P0/P1 ERRORS.md ledger — baked into the scaffolding itself. It is aimed at individual developers and small teams using Claude Code who want better session continuity and AI codebase comprehension without adopting a heavyweight framework. At version 2.2.1 with 89 automated tests, security-hardened dependencies, and an active release cadence, it is production-ready for its defined scope, though its upgrade system is still marked early alpha.

01

Overview

Claude Conductor — Overview

Tagline / Intro (verbatim from README)

A lightweight + modular documentation framework designed for AI-assisted development with Claude Code.

Create a comprehensive, interconnected scaffolded documentation system that helps Claude Code understand and navigate your codebase more effectively, and retain better context.

Origin Story

Claude Conductor was created by Super Basic Studio (a single-person LLC) and first published to GitHub in June 2025. The initial commit message was simply "INIT repo." The project grew organically from a practical need: when Claude Code operates across sessions it loses context about a project's architecture, active tasks, and recent decisions. Rather than relying on ad-hoc CLAUDE.md files that developers wrote from scratch, the author wanted a CLI that could auto-scaffold a structured, AI-optimized documentation suite on demand.

The first notable community contribution came from a Reddit user ("FunnyRocker") who proposed the TASKS.md functionality, which is now acknowledged in the repository's THANKS.md. This reflects the project's grassroots trajectory — it gained traction among "vibe coders" and open-source newcomers before expanding into a more fully featured tool.

Philosophy (verbatim from README — Framework Philosophy section)

  1. Modular > Monolithic - Separate concerns into focused files
  2. Practical > Theoretical - Include real examples and patterns
  3. Maintained > Stale - Regular updates through development
  4. Navigable > Comprehensive - Easy to find what you need

Companion Project

The author explicitly positions Claude Conductor alongside a second project, Claude Anchor (superbasicstudio/claude-anchor):

Conductor is the codebase brain — it documents your project's architecture, APIs, build systems, errors, and development history. It tells Claude what your project is.

Claude Anchor is the behavioral brain — it manages Claude's rules, memory, conversation preferences, and session continuity. It tells Claude how to think and behave.

Together they give Claude full context — what the project is AND how to work on it.

Privacy Commitment

A notable aspect of the project philosophy is an explicit "zero network requests" privacy guarantee:

Claude Conductor is a completely offline tool that:

  • Never collects or transmits any data
  • Makes zero network requests
  • Has no analytics or telemetry
  • Works entirely on your local machine
  • Only reads/writes files you explicitly specify

Maintainer Transparency

The README includes an unusually candid open-source disclaimer:

This is open source software currently maintained by ONE individual in their free time (which isn't much at the moment!)

The project has 3 listed contributors: the maintainer account, the Claude bot (co-authored commits), and Dependabot.

02

Architecture

Claude Conductor — Architecture

Distribution Type

npm-package (published to npmjs as claude-conductor, consumed via npx or global install)

Install Methods

# Recommended: one-shot via npx
npx claude-conductor

# Shorthand alias
npx claude-conduct

# Global install (npm / pnpm / yarn / bun)
npm install -g claude-conductor
pnpm add -g claude-conductor
yarn global add claude-conductor
bun add -g claude-conductor

# Dev dependency added to project
npm install --save-dev claude-conductor

File / Directory Layout (repo root)

superbasicstudio/claude-conductor/
├── bin/
│   └── init.js                  # Single CLI entry point (all commands)
├── templates/
│   ├── CONDUCTOR.md             # Master navigation hub template
│   ├── CLAUDE.md                # AI assistant guidance template
│   ├── ARCHITECTURE.md          # Tech stack / system design template
│   ├── BUILD.md                 # Build & deployment commands template
│   ├── API.md                   # API endpoints template
│   ├── CONFIG.md                # Environment variables template
│   ├── DATA_MODEL.md            # Database schema template
│   ├── DESIGN.md                # Visual design system template
│   ├── UIUX.md                  # UI/UX patterns template
│   ├── TEST.md                  # Testing strategies template
│   ├── CONTRIBUTING.md          # Contribution guidelines template
│   ├── ERRORS.md                # Critical error ledger template
│   ├── TASKS.md                 # Active task management template
│   └── PLAYBOOKS/
│       └── DEPLOY.md            # Deployment procedures template
├── docs/                        # GitHub Pages documentation site
├── test/                        # Jest test suite (89 tests as of v2.2.0)
├── .github/                     # CI workflows, Dependabot config
├── .husky/                      # Pre-commit hooks (lint-staged)
├── CLAUDE.md                    # Conductor's own instance of itself
├── CONDUCTOR.md                 # (Not present in repo root; generated by CLI)
├── JOURNAL.md                   # Conductor project's own dev journal
├── RESPONSE_STYLE_CONFIG.md     # Optional confidence-indicator config
├── package.json
└── package-lock.json

Required Dependencies

  • Node.js >= 20.0.0 (raised from 18.0.0 in v2.2.0 due to Node 18 EOL)
  • Runtime npm packages (from package.json):
    • commander — CLI argument parsing
    • fs-extra — Enhanced file system operations
    • glob — File pattern matching for codebase analysis
    • chalk — Terminal output colorization
  • No Python, Docker, or AI-tool-specific runtime required
  • No internet access required at runtime

Configuration Files Generated

When run, the CLI writes Markdown files directly into the target project directory:

Core set (default npx claude-conductor):

  • CONDUCTOR.md — Master navigation hub
  • CLAUDE.md — AI assistant guidance (auto-skipped if one already exists)
  • ARCHITECTURE.md — System design (auto-populated from codebase scan)
  • BUILD.md — Build/test/deploy commands
  • JOURNAL.md — Dev changelog (first entry created)

Full set (--full flag, adds):

  • API.md, CONFIG.md, DATA_MODEL.md, DESIGN.md, UIUX.md
  • TEST.md, CONTRIBUTING.md, ERRORS.md, TASKS.md
  • PLAYBOOKS/DEPLOY.md

Backup/upgrade artifacts:

  • ./conductor-backup/ — Temporary backup directory during upgrade flow

No .claude/ directory, no agents.md, no command definitions, no settings.json are written.

03

Components

Claude Conductor — Components

Commands

Claude Conductor ships one CLI binary (claude-conductor / claude-conduct) with the following subcommands. These are CLI commands, not Claude Code slash-commands.

Subcommand Purpose
init [target-dir] (default) Scaffold documentation templates; accepts --full, --force, --deepscan, --no-analyze, -y flags
checkup Print a security-audit prompt for the user to paste into Claude Code; checks for exposed secrets, XSS patterns, missing .gitignore entries
backup [target-dir] Step 1 of upgrade: copy JOURNAL.md + CLAUDE.md to ./conductor-backup/
upgrade [target-dir] Step 2 of upgrade: delete all conductor files and reinstall fresh templates (requires --clean flag)
restore [target-dir] Step 3 of upgrade: copy backed-up files back and add an upgrade journal entry

Total CLI subcommands: 5

Skills

(none) — Claude Conductor does not ship Claude Code skills (.claude/skills/ or skills/<name>/SKILL.md).

Subagents

(none) — No named subagents, personas, or role definitions are included.

Hooks

(none) — No Claude Code hooks (PreToolUse, PostToolUse, UserPromptSubmit, SessionStart, Stop, etc.) are defined or installed.

MCP Servers

(none) — No MCP servers are bundled or required.

Scripts / Binaries

Binary Source Description
claude-conductor bin/init.js Primary CLI entry point
claude-conduct (alias, registered in package.json bin) Shorthand alias for the same binary

The binary is a single Node.js script (~750+ lines) implementing all subcommands via the commander library. The codebase analysis (tech stack detection, framework detection, API mapping, component discovery, build script extraction) is all inline in bin/init.js — there are no separate module files.

Template Files (Documentation Framework Core)

While not "commands" or "skills" in the Claude Code sense, the following templates are the deliverable of this framework:

Core templates (always generated):

  1. CONDUCTOR.md — Master navigation hub and journaling rules
  2. CLAUDE.md — AI assistant guidance with Golden Rules, session checklist, task management integration
  3. ARCHITECTURE.md — Tech stack, directory structure, component diagrams
  4. BUILD.md — Build/test/deploy commands
  5. JOURNAL.md — Dev changelog with auto-archive rule at 500 lines

Extended templates (--full flag): 6. API.md — Endpoint documentation with request/response format 7. CONFIG.md — Environment variables and feature flags 8. DATA_MODEL.md — Database schema and data structures 9. DESIGN.md — Visual design system and tokens 10. UIUX.md — Interaction patterns and user flows 11. TEST.md — Testing strategies and coverage goals 12. CONTRIBUTING.md — Contribution guidelines 13. ERRORS.md — P0/P1 critical error ledger with ERR-YYYY-MM-DD-NNN IDs 14. TASKS.md — Active task management with TASK-YYYY-MM-DD-NNN IDs and phase tracking 15. PLAYBOOKS/DEPLOY.md — Deployment procedures and rollback

Total templates: 15 (14 in full set + JOURNAL.md always created)

05

Prompts

Claude Conductor — Prompts & Templates

Claude Conductor's "prompts" are the Markdown templates it writes into user projects. Claude Code reads these files automatically (especially CLAUDE.md) at session start. Below are verbatim excerpts showing the prompting technique.


Excerpt 1: templates/CLAUDE.md — Session Startup Checklist and Golden Rules (verbatim)

This is the core behavioral prompt injected into every project. Note the hard "NEVER" constraints (Golden Rules) — a directive prompting style rather than soft guidance.

## Session Startup Checklist
**IMPORTANT**: At the start of each session, check these items:
1. **Check TASKS.md** - Look for any IN_PROGRESS or BLOCKED tasks from previous sessions
2. **Review recent JOURNAL.md entries** - Scan last 2-3 entries for context
3. **If resuming work**: Load the current task context from TASKS.md before proceeding

---

## GOLDEN RULE — YOU ARE NOT THE PROJECT MANAGER

**NEVER deprioritize, dismiss, or defer any issue the user raises.**

- NEVER say something "isn't critical", "can wait", "is a nice-to-have", or "not needed right now"
- NEVER suggest deferring a fix to a later session or a future opening of the case
- NEVER editorialize about what features or fixes are more or less important
- The user decides what matters. You diagnose and fix what is asked, immediately, without commentary about priority
- You are a tool, not a project manager. No opinions on roadmap, priority, or scheduling

**If a component is broken, diagnose it and fix it. Period.**

---

## GOLDEN RULE — NEVER BLAME HARDWARE FIRST

**When something isn't working, NEVER jump to "hardware issue", "bad wiring", or "damaged component" as a conclusion.**

- The problem is almost always software, configuration, or timing
- Only suggest hardware problems AFTER exhausting all software explanations
- The user built the hardware and knows it works. Trust that.
- Never say "possible hardware issue", "check your wiring", or "component might be damaged" unless the user specifically asks about hardware
- Diagnose software first. Always.

Excerpt 2: templates/CONDUCTOR.md — Journaling Rules and Compaction Rule (verbatim)

CONDUCTOR.md defines the journaling protocol that Claude must follow. The "Compaction Rule" is a built-in context management strategy:

## Continuous Engineering Journal <!-- do not remove -->

Claude, keep an ever-growing changelog in [`JOURNAL.md`](JOURNAL.md).

### What to Journal
- **Major changes**: New features, significant refactors, API changes
- **Bug fixes**: What broke, why, and how it was fixed
- **Frustration points**: Problems that took multiple attempts to solve
- **Design decisions**: Why we chose one approach over another
- **Performance improvements**: Before/after metrics
- **User feedback**: Notable issues or requests
- **Learning moments**: New techniques or patterns discovered

### Journal Format

YYYY-MM-DD HH:MM

[Short Title]

  • What: Brief description of the change
  • Why: Reason for the change
  • How: Technical approach taken
  • Issues: Any problems encountered
  • Result: Outcome and any metrics

[Short Title] |ERROR:ERR-YYYY-MM-DD-001|

  • What: Critical P0/P1 error description
  • Why: Root cause analysis
  • How: Fix implementation
  • Issues: Debugging challenges
  • Result: Resolution and prevention measures

[Task Title] |TASK:TASK-YYYY-MM-DD-001|

  • What: Task implementation summary
  • Why: Part of [Phase Name] phase
  • How: Technical approach and key decisions
  • Issues: Blockers encountered and resolved
  • Result: Task completed, findings documented in ARCHITECTURE.md

### Compaction Rule
When `JOURNAL.md` exceeds **500 lines**:
1. Claude summarizes the oldest half into `JOURNAL_ARCHIVE/<year>-<month>.md`
2. Remaining entries stay in `JOURNAL.md` so the file never grows unbounded

> ⚠️ Claude must NEVER delete raw history—only move & summarize.

Excerpt 3: templates/TASKS.md — Task Context Preservation Template (verbatim)

This template defines the inter-session context handoff format. Each task record is structured to give Claude everything it needs to resume work cold:

# Task Management

## Active Phase
**Phase**: [High-level project phase name]
**Started**: YYYY-MM-DD
**Target**: YYYY-MM-DD
**Progress**: X/Y tasks completed

## Current Task
**Task ID**: TASK-YYYY-MM-DD-NNN
**Title**: [Descriptive task name]
**Status**: PLANNING | IN_PROGRESS | BLOCKED | TESTING | COMPLETE
**Started**: YYYY-MM-DD HH:MM
**Dependencies**: [List task IDs this depends on]

### Task Context
<!-- Critical information needed to resume this task -->
- **Previous Work**: [Link to related tasks/PRs]
- **Key Files**: [Primary files being modified with line ranges]
- **Environment**: [Specific config/versions if relevant]
- **Next Steps**: [Immediate actions when resuming]

### Findings & Decisions
- **FINDING-001**: [Discovery that affects approach]
- **DECISION-001**: [Technical choice made] → Link to ARCHITECTURE.md
- **BLOCKER-001**: [Issue preventing progress] → Link to resolution

### Task Chain
1. ✅ [Completed prerequisite task] (TASK-YYYY-MM-DD-001)
2. 🔄 [Current task] (CURRENT)
3. ⏳ [Next planned task]
4. ⏳ [Future task in phase]

Excerpt 4: Next-Steps Setup Prompts (verbatim from README)

These are the suggested prompts the user pastes into Claude Code after running the CLI — demonstrating Claude Conductor's "scaffold then prompt" pattern:

Quick Setup prompt:

"Please review this codebase and update the CLAUDE.md and CONDUCTOR.md files with the actual project details. Also perform a security health check and list any potential vulnerabilities or concerns (like exposed .env files, API keys in code, missing .gitignore entries, outdated dependencies with known vulnerabilities, or insecure configurations) - just list them as warnings, don't fix anything."

Comprehensive Setup prompt:

"Please thoroughly review this codebase, update CLAUDE.md with project context, and use CONDUCTOR.md as a guide to fill out all the documentation files. Also check for any syntax errors, bugs, or suggestions for improvement. Additionally, perform a comprehensive security health check and list any potential vulnerabilities or concerns (like exposed .env files, API keys in code, missing .gitignore entries, outdated dependencies with known vulnerabilities, insecure configurations, or other security best practice violations) - just list them as warnings, don't fix anything."

Excerpt 5: templates/ERRORS.md — P0/P1 Error Ledger Structure (verbatim)

# Critical Error Ledger <!-- auto-maintained -->

## Schema
| ID | First seen | Status | Severity | Affected area | Link to fix |
|----|------------|--------|----------|---------------|-------------|

## Active Errors
[New errors added here, newest first]

## Resolved Errors
[Moved here when fixed, with links to fixes]

---

## Error Severity Definitions

### P0 - Critical (System Down)
- Complete service outage
- Data loss or corruption
- Security breach
- Payment system failure

### P1 - Major (Degraded Service)
- Core functionality broken
- Significant performance issues
- Authentication failures
- Data sync problems

Prompting Technique Summary

Claude Conductor uses three distinct prompting patterns:

  1. Directive constraints with "NEVER" — Hard prohibitions in GOLDEN RULE sections prevent Claude from making autonomous priority decisions or blaming hardware. These are unconditional bans, not preferences.

  2. Structured tagging protocol — Journal entries embed machine-parseable tags (|ERROR:ERR-ID|, |TASK:TASK-ID|) that allow Claude to cross-reference documentation files without ambiguity, enabling reliable context recovery in future sessions.

  3. Session startup checklist — Rather than relying on Claude to remember context, the CLAUDE.md template instructs Claude to perform an explicit three-step state check at the start of every session (TASKS.md → JOURNAL.md → resume), making context recovery a mandatory procedure.

09

Uniqueness

Claude Conductor — Uniqueness & Variants

What Does This Do That No Other Seed Framework Does?

Claude Conductor's distinctive combination is: a zero-dependency offline npm CLI that auto-analyzes a codebase and writes a full cross-linked documentation suite, with a structured error ledger (P0/P1 with IDs), a compacting journal, and a task context-preservation format — all as plain Markdown, with no external services.

Specifically:

  • The ERRORS.md ledger with ERR-YYYY-MM-DD-NNN IDs linked to journal entries is not found in other frameworks analyzed in this research set.
  • The TASKS.md inter-session handoff format with explicit TASK-YYYY-MM-DD-NNN IDs, task chains, and finding/decision logs is a distinctive context-recovery primitive.
  • The "500-line compaction rule" for JOURNAL.md (Claude archives, never deletes) is an explicit compaction strategy written into the template itself.
  • The CLI's --deepscan mode, which auto-populates ARCHITECTURE.md with detected frameworks, API endpoints, React/Vue components, and build scripts, is a generation feature not found in pure template bundles.

What Does It Explicitly DROP That Others Have?

  • No slash-commands — Does not create .claude/commands/ entries. Users cannot invoke any workflow step as /command.
  • No skills — Does not ship Claude Code skills.
  • No hooks — No PreToolUse, PostToolUse, or other event hooks.
  • No agents or subagents — No multi-agent orchestration.
  • No MCP integration — No MCP servers, no tool definitions.
  • No spec/plan/tasks generation workflow — No "Brainstorm → Spec → Plan → Tasks → Implement" loop. It scaffolds docs but does not drive a feature development lifecycle.
  • No TDD enforcement — TEST.md is a documentation template; tests are not generated or run.
  • No AI provider coupling — The CLI itself makes zero AI calls; it is a static file generator.

One-Sentence Positioning

Claude Conductor is a doc-scaffold-first framework: it believes the most leveraged investment in AI-assisted development is giving the AI a well-structured, maintained documentation system to navigate, not workflow automation or prompt engineering.

Failure Modes / Criticisms

  • Single maintainer risk — Explicitly acknowledged in the README. One person, limited free time. The project could go dormant without notice.
  • Manual discipline required — The journaling, task management, and error ledger only work if Claude (and the developer) actually maintain them. The CLI creates empty templates; it cannot enforce ongoing hygiene.
  • CLAUDE.md collision — If the project already has a CLAUDE.md (common in Claude Code projects), the CLI skips creating one and asks the user to manually add a ## Journal Update Requirements block. This is a friction point.
  • Upgrade system marked alpha — The backup/upgrade/restore flow is explicitly labeled "EARLY ALPHA FEATURE - USE WITH EXTREME CAUTION" in the README as of v2.2.1.
  • No verification that Claude reads the docs — The framework assumes Claude Code will read the generated files, but there is no hook or mechanism to confirm or enforce this.

Variant / Derivative Repos Named "Conductor"

The primary repo is superbasicstudio/claude-conductor. Several other repos share the name or concept:

1. lackeyjb/claude-conductor

  • A fork or derivative targeting Gemini CLI rather than Claude Code.
  • README states it is "inspired by Gemini Conductor" (a separate project in the Gemini ecosystem).
  • Relationship: Parallel derivative using the "conductor" naming convention; not a fork of superbasicstudio's repo.

2. fcoury/conductor

  • A different tool (not Claude-specific). Purpose: unknown from available data.
  • Relationship: Name collision only; unrelated to AI documentation frameworks.

3. MZWASHERE/claude-conductor

  • Listed as a fork of superbasicstudio/claude-conductor (one of the 22 forks).
  • No distinguishing features documented.

4. rbarcante/claude-conductor

  • Listed as a fork of superbasicstudio/claude-conductor.
  • No distinguishing features documented.

5. microsoft/conductor

  • Microsoft's "Conductor" is an internal orchestration/workflow tool (not AI documentation).
  • Relationship: Name collision only; entirely different domain.

6. MadAppGang/claude-code (Conductor plugin)

  • A Claude Code plugin that includes a "Conductor" component for multi-agent orchestration.
  • This is a different architecture entirely (Claude Code plugin, not npm CLI, multi-agent).
  • Relationship: Name-inspired but different paradigm; targets Claude Code's plugin system rather than the documentation layer.

Summary Table

Repo Type Relationship to Primary
superbasicstudio/claude-conductor npm CLI, doc scaffold PRIMARY
lackeyjb/claude-conductor Gemini CLI variant Inspired-by, different tool
fcoury/conductor Unknown Name collision
MZWASHERE/claude-conductor Fork Fork of primary
rbarcante/claude-conductor Fork Fork of primary
microsoft/conductor Enterprise orchestration Name collision
MadAppGang/claude-code (Conductor plugin) Claude Code plugin Name-inspired, different paradigm
04

Workflow

Claude Conductor — Workflow

Workflow Phases

Claude Conductor itself does not prescribe a multi-phase development workflow. Instead, it scaffolds a documentation infrastructure that supports any workflow. The phases below are the setup phases described in CONDUCTOR.md:

Phase Duration Artifact
Phase 1: Initial Setup 30–60 min CLAUDE.md with Critical Context section populated
Phase 2: Core Documentation 2–4 hours All .md files created, keywords added, cross-links established, ERRORS.md schema configured
Phase 3: Optimization 1–2 hours Task Templates in CLAUDE.md, ASCII flow diagrams in ARCHITECTURE.md, Anti-Patterns documented
Phase 4: First Journal Entry Minutes JOURNAL.md initial entry documenting the setup

After initialization, the ongoing session workflow implied by the templates is:

  1. Session Start — Claude reads TASKS.md for IN_PROGRESS/BLOCKED tasks, reads last 2–3 JOURNAL.md entries
  2. Active Work — Claude implements features/fixes
  3. Task Completion — Claude updates TASKS.md status, generates JOURNAL.md entry with |TASK:ID| tag, archives task to TASKS_ARCHIVE/YYYY-MM/
  4. Error Encounter — Claude adds ERR-YYYY-MM-DD-NNN entry to ERRORS.md, creates linked JOURNAL.md entry
  5. Journal Maintenance — When JOURNAL.md exceeds 500 lines, Claude moves oldest half to JOURNAL_ARCHIVE/YYYY-MM.md

Human Approval Gates

The CLI itself has one user-facing prompt: when --force is not passed and files already exist, the CLI asks for confirmation before overwriting. The --yes / -y flag bypasses this.

No explicit human approval gates are defined within the documentation framework's workflow instructions. The TASKS.md template defines task states (PLANNING, IN_PROGRESS, BLOCKED, TESTING, COMPLETE) but does not mandate user sign-off between transitions.

TDD Enforcement

No. The framework does not enforce TDD. A TEST.md template is included in the full set (covering test stack, running commands, coverage goals, common patterns), but it is a documentation template only — no test generation, no pre-commit test gates, and no mention of red/green/refactor in the workflow instructions.

Multi-Agent Execution

No. Claude Conductor is single-agent. All documentation and development work is performed by one Claude Code session. There is no orchestration layer, no subagent dispatch, and no parallelism model.

Git Worktrees / Isolated Workspaces

No. No worktree or branch isolation is part of the framework.

Spec Format

The framework does not use a formal spec format (no Gherkin, YAML spec, or JSON schema). Documentation is authored in plain Markdown with a set of structural conventions (section headers, checkbox lists, line-number anchors, keyword sections).

Files Generated Per Feature

Claude Conductor does not have a per-feature artifact model. Instead, the following files are maintained continuously and updated across all features:

  • JOURNAL.md — Updated after every significant change
  • TASKS.md — Updated when a task changes state
  • ARCHITECTURE.md — Updated when structure changes significantly
  • ERRORS.md — Updated when P0/P1 errors occur or resolve
  • TASKS_ARCHIVE/YYYY-MM/TASK-ID.md — Archived when a task completes

There is no per-feature spec.md, plan.md, or tasks.md generation.

06

Memory Context

Claude Conductor — Memory & Context

Memory Model

file-based

Claude Conductor's entire memory system is plain Markdown files written to disk. There is no database, no vector store, no MCP bridge, and no in-process state beyond what the CLI reads during a single run.

Persistence Scope

project

All generated files (JOURNAL.md, TASKS.md, ERRORS.md, etc.) live in the target project directory. They persist across sessions because they are version-controlled files, not session data. There is no global memory and no cross-project memory.

Context Compaction Strategy

The framework defines an explicit compaction rule in CONDUCTOR.md (verbatim):

When JOURNAL.md exceeds 500 lines:

  1. Claude summarizes the oldest half into JOURNAL_ARCHIVE/<year>-<month>.md
  2. Remaining entries stay in JOURNAL.md so the file never grows unbounded ⚠️ Claude must NEVER delete raw history—only move & summarize.

This is a manual compaction protocol — the CLI does not perform compaction automatically. Claude Code is instructed (via CONDUCTOR.md) to execute the archive step when the threshold is crossed. Similarly, completed tasks are archived to TASKS_ARCHIVE/YYYY-MM/TASK-ID.md by Claude.

Cross-Session Handoffs

Yes — cross-session handoff is the primary design goal of the framework.

The mechanism is:

  1. TASKS.md records the current task's ID, status, key files with line ranges, environment notes, next steps, findings, and decisions — everything Claude needs to resume cold.
  2. JOURNAL.md records a timestamped history with task IDs (|TASK:ID| tags) and error IDs (|ERROR:ID| tags).
  3. CLAUDE.md's Session Startup Checklist instructs Claude to read TASKS.md and the last 2–3 JOURNAL.md entries at the start of every session.
  4. If Claude Code is reinstalled or the session is fully reset, the JOURNAL.md tag search (search JOURNAL.md for |TASK:) can recover the full history of completed tasks.

Reference to Memory Concepts

The README and templates use the following terms:

  • "Development Journal" — the JOURNAL.md continuous changelog
  • "Context preservation" — TASKS.md's stated purpose; phrase used in README feature list
  • "Context Recovery" — described in CLAUDE.md: "Search JOURNAL.md for |TASK: to see all completed tasks over time"
  • "Clean Handoffs" — CLAUDE.md: "TASKS.md always shows what needs to be resumed or completed"

There is no reference to "memory bank", "knowledge base", or vector/embedding concepts.

RESPONSE_STYLE_CONFIG.md (Optional Confidence System)

A supplementary file (RESPONSE_STYLE_CONFIG.md) ships in the repo root (not as a template). It defines an optional add-on for CLAUDE.md that configures confidence-indicator display:

## Response Style Settings
- **CONFIDENCE_INDICATORS**: enabled
- **TONE_CONTROL**: strict

When enabled, Claude shows visual confidence bars (e.g., [███████░░░] 70%) and avoids premature "DONE!" declarations. This is not memory per se but affects how Claude reports its certainty about context-dependent tasks.

07

Target Tools

Claude Conductor — Target Tools

Officially Supported Tools

Claude Code (primary)

Evidence: The framework is explicitly named "Claude Conductor — for Claude Code command line." Every template references Claude Code behavior. The README states: "A lightweight + modular documentation framework designed for AI-assisted development with Claude Code." CLAUDE.md is the standard Claude Code context file. The CLI binary description is "Documentation framework for AI-assisted development."

Install path: npx claude-conductor run from the project root. The generated files (CLAUDE.md, CONDUCTOR.md, etc.) are placed in the project directory where Claude Code reads them automatically.

Compatibility: Requires Node.js >= 20.0.0 on the developer's machine. No Claude Code version constraint is stated; the framework relies on Claude Code's standard behavior of reading CLAUDE.md at session start.


No Other Tools Listed

The README, templates, and CLI source make no mention of, and provide no install paths for:

  • Cursor
  • Codex / OpenAI Codex
  • Aider
  • Cline
  • GitHub Copilot
  • Opencode
  • Goose
  • Windsurf
  • Gemini CLI
  • Roo
  • Kilo
  • Qwen
  • Jules
  • Continue
  • Any other AI coding assistant

Compatibility Notes

While the generated Markdown files are generic enough that any AI coding tool that reads project documentation (e.g., Cursor's .cursorrules pattern) could theoretically benefit from them, Claude Conductor makes no claim to support these tools and provides no configuration targeting them.

The companion project, Claude Anchor (superbasicstudio/claude-anchor), also targets Claude Code specifically and is described as complementary to Claude Conductor.

08

Signals

Claude Conductor — Signals

GitHub Stars

367 (as of 2026-05-26, via gh api /repos/superbasicstudio/claude-conductor)

Star history chart is embedded in the README and shows a sharp spike consistent with a viral Reddit or social post shortly after the June 2025 launch.

Last Commit Date

2026-05-17 (commit 4bc2332: chore(release): 2.2.1 — commander 14 compatibility fix)

Number of Contributors

2 human contributors (superbasicstudio + dependabot bot). The "claude" account listed in contributor graphs is the Anthropic bot that co-authored commits (Co-Authored-By: Claude Opus 4.6), not a human contributor.

From gh api /repos/superbasicstudio/claude-conductor/contributors:

  • superbasicstudio: 56 commits
  • dependabot[bot]: 1 commit

Forks

22 forks

Open Issues

10 open issues at time of analysis

Pull Requests

10 open pull requests at time of analysis

Reddit / HN Sentiment

The project's THANKS.md credits a Reddit user ("FunnyRocker") for proposing the TASKS.md feature, confirming Reddit community engagement. The README explicitly references "vibe coders" as a target audience, suggesting awareness of the Reddit/vibe-coding community.

Specific Reddit thread sentiment: unknown (no wave-2b-reddit-discovery.md quotes available for this project in the research index).

HN sentiment: unknown

Maintainer Status

Active — The most recent commit is from 2026-05-17 (9 days before this analysis). The project has had 8 releases (v1.0.x through v2.2.1), consistent CI/CD via GitHub Actions, Dependabot configured for daily dependency checks, and a 57-commit history spanning from June 2025 to May 2026. The maintainer's own README note ("maintained by ONE individual in their free time") suggests active but potentially constrained bandwidth.

Release History

8 tagged releases: v2.2.1 (latest), v2.2.0 (security hardening + Node 20), v2.1.0 (linting/formatting/CI), v2.0.x, v1.3.0, and earlier 1.x versions.

Related frameworks

same archetype · same primary tool · same memory type

BMAD-METHOD ★ 48k

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

Agent OS ★ 4.6k

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

Spec-Driver (Greenfield Spec-Driven Development) ★ 25

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

Anthropic Knowledge Work Plugins ★ 16k

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

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

mini-coding-agent ★ 882

A single-file zero-dependency Python coding agent that demonstrates the six core components of coding agents for educational…