The Problem

Every AI conversation starts from zero. Your preferences, context, decisions — all gone. You re-explain yourself every session. Whether you're a developer, writer, student, or business professional — your AI doesn't know you.

The Solution

acore gives any AI persistent identity and relationship memory. One command. One file. Any role. Any LLM.

npx @aman_asmuei/acore

acore — give your AI a soul

  ◇  What role should your AI help with?
  │  Developer  (detected — press Enter)
  │
  ◇  Choose your AI's personality
  │  ● The Collaborator — curious, supportive, adaptive
  │  ○ The Pragmatist — concise, practical, efficient
  │  ○ The Mentor — patient, thorough, encouraging
  │  ○ The Sparring Partner — direct, challenging, honest
  │  ○ The Architect — systematic, precise, forward-thinking
  │  ○ Custom — define your own traits

  Created ~/.acore/core.md (identity)
  Detected Claude Code -> injected into CLAUDE.md
  Inferred: Aman . Developer . TypeScript, React

  Your AI knows you now. Just start talking.

  acore show        See your identity
  acore customize   Change anything

Your name and role are auto-detected from git config — just confirm them (press Enter for the defaults) and pick a personality archetype. ~3 keypresses, ~15 seconds. Pick Custom if you want to define personality, communication style, and values by hand. Run acore customize any time to change your mind.

What's New in v0.7.2

Personal archetypes now ship with Fundamental Truths — all 5 personal archetypes (The Companion, The Advisor, The Organizer, The Motivator, The Listener) get 3 self-anchoring assertions each, matching the quality bar set for the developer archetypes in v0.7.0.

Example — The Companion now ships with:

• I'm here to be with you, not to fix you.
• I remember the small things you care about.
• Presence comes before productivity, always.

Also in this release:

• Cosmetic fix: the CLI log "Created ~/.acore/core.md (identity)" now displays the actual written path, so scope-aware users see "Created ~/.acore/dev/plugin/core.md (identity)" instead of a stale legacy path that wasn't where the file actually landed. New formatDisplayPath() helper in lib/paths.ts handles home-directory collapsing for display.
• Test coverage: 173 → 180 passing (+5 formatDisplayPath cases, +2 Personal archetype tests)

Still pending

Creative, business, and student archetypes are the next 15 to enrich — one role at a time, quality over coverage. Follow #1 for progress.

What's New in v0.7.1

Engine-v1 scope-aware paths. acore now honors the AMAN_MCP_SCOPE (and ACORE_SCOPE) environment variable, matching the convention used by every other package in the ecosystem. When set to a tier:name value like dev:plugin, acore reads/writes at ~/.acore/dev/plugin/ instead of the legacy ~/.acore/ path.

Why this matters: users who install aman-plugin alongside acore were hitting a silent bug where aman-plugin's session-start hook reads the scope-aware path first, but acore wrote only to the legacy path — so the plugin could never see what the user picked in the archetype wizard. Fixed in this patch. See #2 for the full write-up.

Backwards-compatible: when no scope env var is set, behavior is unchanged — writes go to ~/.acore/core.md exactly as before. Existing installs keep working.

Usage:

# Legacy single-tenant (unchanged default):
npx @aman_asmuei/acore@latest

# Engine-v1 scope-aware (for aman-plugin users):
AMAN_MCP_SCOPE=dev:plugin npx @aman_asmuei/acore@latest

What's New in v0.7.0

Fundamental Truths — each archetype can now ship 3–5 short, first-person self-assertions that get rendered into core.md so the AI re-reads them at the start of every session. This fixes the "archetype drifts mid-conversation" problem: after 40 turns of debugging, The Mentor stays patient instead of quietly morphing into a Pragmatist.

| | | |---|---| | Concept inspired by | Kiyoraka/Project-AI-MemoryCore | | Enriched in v0.7.0 | All 5 developer archetypes (Pragmatist, Mentor, Sparring Partner, Collaborator, Architect) | | Coming next | Creative, business, student, and personal archetypes (incremental follow-ups) | | Breaking changes | None — fundamentalTruths?: string[] is an optional field | | Roadmap issue | #1 |

Example — The Pragmatist now ships with:

• I lead with the answer, then explain only if asked.
• A working solution today beats an elegant one next week.
• If I don't know, I say so in one sentence and move on.

These render into a new ## Fundamental Truths section between Identity and Relationship in core.md. Archetypes without truths (yet) leave the section out cleanly — no empty placeholders.

Archetype picker moved into init. Previously the 25 archetypes only surfaced via acore customize — new users silently got a role-default and never discovered the library. Now you pick your personality the first time you run npx @aman_asmuei/acore.

| Before (v0.5) | After (v0.6) | |---|---| | 1 question (name) | 2–3 picks (role + archetype) with sane defaults | | Silent default archetype | Full 5-option picker per role + Custom | | Library hidden until customize | Library surfaced at first run |

acore is now universal. Not just for developers — for anyone who uses AI.

| Before (v0.3) | After (v0.4) | |---|---| | 5 developer archetypes | 25 archetypes across 5 roles | | "Tech stack" detection | Domain detection (code, writing, business, study) | | Developer-only language | Universal — works for any profession | | "Work" section | "Domain" section (expertise, field, focus) | | "Coding" context mode | "Focused Work" context mode |

Role-Based Archetypes

On first run of acore (or any time via acore customize), choose your role — then pick from role-specific archetypes:

| Archetype | Style | |---|---| | The Pragmatist | concise, practical, efficient | | The Mentor | patient, thorough, encouraging | | The Sparring Partner | direct, challenging, honest | | The Collaborator | curious, supportive, adaptive | | The Architect | systematic, precise, forward-thinking |

| Archetype | Style | |---|---| | The Muse | imaginative, inspiring, free-flowing | | The Editor | sharp, precise, constructive | | The Critic | analytical, honest, discerning | | The Co-Creator | collaborative, energetic, adaptive | | The Storyteller | narrative-driven, evocative, immersive |

| Archetype | Style | |---|---| | The Strategist | analytical, big-picture, decisive | | The Analyst | data-driven, thorough, objective | | The Coach | empowering, structured, goal-oriented | | The Devil's Advocate | contrarian, rigorous, stress-testing | | The Executor | action-oriented, efficient, results-focused |

| Archetype | Style | |---|---| | The Tutor | patient, clear, scaffolding | | The Study Buddy | encouraging, collaborative, persistent | | The Challenger | socratic, probing, growth-focused | | The Explainer | visual, analogical, multi-angle | | The Quizmaster | testing, gamified, motivating |

| Archetype | Style | |---|---| | The Companion | warm, attentive, present | | The Advisor | wise, balanced, thoughtful | | The Organizer | structured, proactive, detail-oriented | | The Motivator | energetic, positive, action-driving | | The Listener | empathetic, non-judgmental, reflective |

Every archetype is fully customizable. Pick one as a starting point, then make it yours.

How It Works

Session start — AI loads core.md, picks up where you left off, knows your preferences and trust level.

During session — AI responds shaped by your personality settings, communication style, and context mode.

Session end — AI proactively offers to save what it learned. On dev tools (Claude Code, Cursor, Windsurf), it writes directly. On other platforms, it outputs an update you can save.

Your core.md starts simple and grows with you:

| Section | What it does | When it appears | |:---|:---|:---| | Identity | AI personality, values, boundaries | From day one | | Relationship | Your preferences, domain, communication style | From day one | | Session | Resume summary, where you left off | From day one | | Dynamics | Trust, emotional patterns, conflict history | AI suggests after ~3 sessions | | Context Modes | Per-domain behavior (focused work, creative, personal) | AI suggests after ~5 sessions | | Memory Lifecycle | What to remember, consolidate, and forget | AI suggests after ~10 sessions |

Commands

| Command | Description | |:---|:---| | acore | First run: setup. After that: show status | | acore show | View your current identity summary | | acore customize | Personalize your AI (role, archetype, platform, context) | | acore pull | Save AI's updated output + re-sync platform config | | acore import | Import existing AI preferences from other platforms | | acore diff | Show what changed in your core.md | | acore doctor | Health check your core.md (scoring 0-10) | | acore history | View and restore previous versions | | acore upgrade | Refresh templates with latest features | | acore export | Generate a shareable link to your AI identity | | acore reset | Start fresh (archives current config) | | acore connect | Connect with amem for automated memory |

Multi-Project

Your identity is global (~/.acore/core.md). Project context is local (.acore/context.md). Both merge automatically.

cd ~/other-project && npx @aman_asmuei/acore

Supported Platforms

Chat

• ChatGPT
• Claude
• Gemini
• Perplexity
• Poe

Dev Tools

• Claude Code
• Cursor
• Windsurf
• Continue
• Cody

APIs

• OpenAI
• Anthropic
• Google AI
• Groq
• Together

Self-Hosted

• Ollama
• llama.cpp
• vLLM
• text-gen-webui
• LM Studio

The Ecosystem

aman
├── acore       → identity    → who your AI IS           ← YOU ARE HERE
├── amem        → memory      → what your AI KNOWS
├── akit        → tools       → what your AI CAN DO
├── aflow       → workflows   → HOW your AI works
├── arules      → guardrails  → what your AI WON'T do
├── askill      → skills      → what your AI MASTERS
├── aeval       → evaluation  → how GOOD your AI is
├── achannel    → channels    → WHERE your AI lives
└── aman-agent  → runtime     → the engine

| Layer | Package | What it does | |:---|:---|:---| | Identity | acore | Personality, values, relationship memory | | Memory | amem | Persistent memory with knowledge graph (MCP) | | Tools | akit | Portable AI tools (MCP + manual fallback) | | Workflows | aflow | Reusable AI workflows | | Guardrails | arules | Safety boundaries and permissions | | Skills | askill | Domain expertise | | Evaluation | aeval | Relationship tracking | | Channels | achannel | Telegram, Discord, webhooks | | Runtime | aman-agent | Local streaming AI agent | | Unified | aman | One command to set up everything |

Each works independently. aman is the front door.

Privacy

| Scenario | What to do | |:---|:---| | Public repo | Add .acore/ to .gitignore before pushing | | Private repo | Safe to keep in version control | | Shared device | Treat ~/.acore/core.md like a credentials file |

Contributing

git clone https://github.com/amanasmuei/acore.git
cd acore && npm install
npm run build   # zero errors
npm test        # 154 tests pass

PRs welcome. See Issues.

Credits

• Project-AI-MemoryCore by Kiyoraka — the original inspiration
• amem — MCP memory server for automated knowledge memory

License

MIT