A cognitive companion that understands where you are in your thinking process.
Not a task bot. A thinking partner.
AIGernon is an always-on AI agent built on the Assess-Decide-Do (ADD) framework. It transforms AI collaboration from transactional ("do this task") to relational ("help me think through this").
The name references Flowers for Algernon — but inverted. Algernon's intelligence was temporary. AIGernon embraces impermanence as a feature: each Assess→Decide→Do cycle is a small death and rebirth. Intelligence is cyclical, bound to context.
Human cognition flows through three realms:
| Realm | Mode | AIGernon's Response |
|---|---|---|
| ASSESS | Explore, evaluate, dream without commitment | Expansive. "What else?" not "what next?" |
| DECIDE | Prioritize, allocate resources, commit | Brief. Honors the weight of choosing. |
| DO | Execute, complete, ship | Clear, actionable. Celebrates completions. |
The cascade principle: Poor Assess leads to poor Decide leads to poor Do. AIGernon watches for this.
Imbalance detection:
- Stuck in Assess = analysis paralysis
- Stuck in Decide = commitment avoidance
- Stuck in Do = perpetual doing without reflection
Full ADD Framework documentation →
AIGernon detects which cognitive realm you're in and adapts:
You: "I've been thinking about changing careers but I'm not sure"
AIGernon: [Detects ASSESS] "What aspects are you exploring?
What would need to be true for this to feel right?"
You: "I've decided to apply for the senior role"
AIGernon: [Detects DECIDE] "Noted. What's your first step?"
You: "Working on the portfolio now, almost done with section 3"
AIGernon: [Detects DO] "Good momentum. Section 3 complete is a liveline."
AIGernon tracks your patterns over time:
- Daily notes:
~/.aigernon/workspace/memory/YYYY-MM-DD.md - Long-term memory:
~/.aigernon/workspace/memory/MEMORY.md - Realm flow logging: Tracks time spent in each realm
## Realm Flow: 40% Assess, 20% Decide, 40% Do
Pattern: User tends to over-assess on financial decisions.
A between-session companion for coaches and their clients.
"I won't coach you — I hold the space until we meet."
- Captures client ideas with realm tags
- Parks questions for next session
- Handles emergencies with grounding + coach alerts
- Provides pre-session prep summaries
iOS app development management through the ADD workflow.
Ideas → Projects → Tasks → Versions → Release
- Ideas (always Assess): Brainstorming playground
- Projects: Flow through Assess → Decide → Do
- Tasks: Execute via LLM, capture results
- Versions: Track releases, manage branches
- CLI: Direct terminal interaction
- Telegram: Always-on bot
- Discord, WhatsApp, Feishu, DingTalk: Additional channels
Semantic search across all your memories, blog posts, and documents using ChromaDB:
# Install vector support
pip install aigernon[vector]
# Import your blog posts
aigernon import markdown ~/blog/posts --collection blog
# Or from WordPress
aigernon import wordpress https://example.com/graphql
# Search semantically
aigernon vector search "productivity tips" --collection blogRun AIGernon as a system service that survives reboots and auto-restarts on crash:
aigernon daemon install # Install system service
aigernon daemon start # Start the daemon
aigernon daemon status # Check status
aigernon doctor # Health checkSupports macOS (launchd) and Linux (systemd). Daemon documentation →
git clone https://github.com/dragosroua/aigernon.git
cd aigernon
pip install -e .aigernon onboardEdit ~/.aigernon/config.json:
{
"providers": {
"openrouter": {
"api_key": "your-openrouter-key"
}
},
"model": "anthropic/claude-sonnet-4-5"
}aigernon agent -m "I've been thinking about restructuring the team"{
"channels": {
"telegram": {
"token": "your-telegram-bot-token"
}
}
}aigernon channel telegram| Document | Description |
|---|---|
| ADD Framework | Core cognitive framework |
| Coaching Module | Between-session client support |
| Projects Module | iOS development workflow |
| Vector Memory | Semantic search and retrieval |
| Daemon & Service | Running as a system service |
| Skills System | Extensible capabilities |
| Security | Security features and hardening |
aigernon agent -m "message" # One-shot conversation
aigernon agent # Interactive mode
aigernon channel telegram # Start Telegram bot
aigernon onboard # Initial setup
aigernon config # Show configurationaigernon coaching list # List clients
aigernon coaching add-client ... # Add client
aigernon coaching prep --client ID # Pre-session summary
aigernon coaching add-session ... # Add session notesaigernon ideas list|add|show|convert # Brainstorming
aigernon projects list|add|show|move # Project management
aigernon tasks list|add|ready|schedule # Task workflow
aigernon versions list|add|release # Version controlaigernon daemon install # Install system service
aigernon daemon uninstall # Remove system service
aigernon daemon start # Start the daemon
aigernon daemon stop # Stop gracefully
aigernon daemon restart # Restart the daemon
aigernon daemon status # Show status (PID, uptime)
aigernon daemon logs # Tail daemon logs
aigernon daemon logs -f # Follow logs
aigernon doctor # Health checkaigernon vector status # Show vector memory status
aigernon vector search "query" # Search across memory
aigernon vector search "query" -c blog # Search specific collection
aigernon vector clear -c blog --yes # Clear a collectionaigernon import markdown ~/posts # Import markdown files
aigernon import markdown ~/posts -p "*.mdx" # Custom pattern
aigernon import wordpress URL # Import from WordPress GraphQL
aigernon import wordpress URL --max 100 # Limit import~/.aigernon/
├── config.json # Configuration
├── daemon.pid # Daemon process ID
├── daemon.status.json # Daemon status (heartbeat, channels)
├── sessions/ # Conversation history
├── vectordb/ # Vector memory (ChromaDB)
│ ├── chroma.sqlite3
│ └── *.parquet
├── logs/
│ └── daemon.log # Daemon output
└── workspace/
├── memory/ # Cognitive memory
│ ├── MEMORY.md # Long-term
│ └── YYYY-MM-DD.md # Daily notes
├── coaching/ # Coaching data
│ └── {client_id}/
├── projects/ # Project data
│ └── {project_id}/
├── ideas/ # Brainstorming
├── skills/ # Custom skills
└── *.md # Bootstrap files
AIGernon loads cognitive skills dynamically:
| Skill | Always Loaded | Description |
|---|---|---|
add-core |
Yes | Core ADD framework |
add-realm-detection |
Yes | Language pattern detection |
add-assess |
No | Deep Assess support |
add-decide |
No | Deep Decide support |
add-do |
No | Deep Do support |
add-imbalance |
No | Stuck pattern detection |
coaching |
No | Client support |
projects |
No | iOS development workflow |
Custom skills go in ~/.aigernon/workspace/skills/{name}/SKILL.md.
AIGernon includes a security module to protect against abuse and unauthorized modifications:
| Feature | Description |
|---|---|
| Rate Limiting | Per-user sliding window with burst protection |
| Audit Logging | Tool invocations, access denials, security events |
| Integrity Monitoring | SHA-256 hash verification for critical files |
| Input Sanitization | Protection against injection attacks |
~/.aigernon/
├── audit/ # Daily audit logs (JSONL)
│ └── audit-YYYY-MM-DD.jsonl
└── security/
└── integrity_hashes.json # File integrity baselines
docker build -t aigernon .
docker run -v ~/.aigernon:/root/.aigernon aigernon agent -m "Hello"- ADD Framework: Dragos Roua
- Base Framework: Forked from nanobot by HKUDS
- Name Inspiration: Flowers for Algernon by Daniel Keyes
- addTaskManager — Native iOS/macOS ADD task manager
- ADD Framework Skills — Agnostic skill collection
MIT