dennisthiessen/claude-inner-dialog
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add missing reference files and streamline documentation

Browse Source 
- Add missing modality files: lifespan-integration, somatic-experiencing, psychodynamic
- Add missing persona file: grounded-real
- Refactor docs to reduce overlap and clarify focus:
  - README: overview and why (removed install steps, links to other docs)
  - GETTING-STARTED: step-by-step setup only
  - SECURITY: privacy options with friendlier language
- Simplify language for non-technical audience
- Remove duplicate FAQ, crisis resources, and privacy explanations

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Anthony Taglianetti
2026-02-01 21:27:44 -08:00
parent 301dfd3c1e
commit 783fd4a5ae
7 changed files with 538 additions and 443 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+75 -165
View File
@@ -1,8 +1,8 @@
# AI Therapy Starter Kit
A local-first, privacy-focused toolkit for AI-assisted therapy and self-reflection.
A private, persistent AI therapist that lives on your computer.
Your sessions stay on your computer. Your data is yours. Works with Claude, GPT, or any LLM.
Your sessions stay local. Your AI remembers everything. Works with Claude, GPT, or any AI.
---
@@ -10,10 +10,10 @@ Your sessions stay on your computer. Your data is yours. Works with Claude, GPT,
If you've used ChatGPT or Claude for emotional support, you've probably hit these problems:
- **Sessions disappear.** Start a new chat, lose all context. Copy-paste to Google Docs? That gets old fast.
- **Privacy concerns.** Your deepest thoughts, stored on someone else's servers, potentially reviewed by humans.
- **Platform lock-in.** Build a relationship with 4o, then OpenAI deprecates it. Start over.
- **Inconsistent quality.** Sometimes helpful, sometimes reads from a script. No therapeutic framework.
- **Sessions disappear.** Start a new chat, lose all context.
- **Privacy concerns.** Your deepest thoughts on someone else's servers.
- **Platform lock-in.** Build a relationship with one AI, then it changes or goes away.
- **Inconsistent quality.** Sometimes helpful, sometimes generic.
This toolkit solves all of them.
@@ -21,58 +21,25 @@ This toolkit solves all of them.
## What You Get
**Persistent memory** — Sessions and insights accumulate in local markdown files. Your AI therapist remembers everything, across sessions, forever.
**Your AI remembers** — Sessions and insights accumulate in files on your computer. Your AI therapist remembers everything, forever.
**Real privacy** — Files live on your machine. Optional encryption (AES-256) for sensitive data. Use a local LLM for maximum privacy.
**Real privacy** — All files stay on your machine. Optional password protection for shared computers.
**Portable** — Works with Claude Code, ChatGPT API, or any LLM that can read files. Switch providers without losing your history.
**Portable** — Works with Claude, ChatGPT, or any AI that can read files. Switch anytime without losing your history.
**Evidence-based framework** — Built-in CBT, ACT, DBT, Somatic Experiencing, and Lifespan Integration. Structured therapeutic approaches, not just vibes.
**Evidence-based** — Built-in therapeutic approaches (CBT, ACT, DBT, and more). Real frameworks, not just conversation.
**Customizable** — Choose your therapist's communication style. Add modalities. Make it yours.
**Customizable** — Choose your therapist's name and communication style. Make it yours.
---
## Quick Start
## Get Started
**New to Claude Code or terminal-based AI tools?** See the [Getting Started Guide](docs/GETTING-STARTED.md) for step-by-step instructions.
**Ready to set up your AI therapist?** See the [Getting Started Guide](docs/GETTING-STARTED.md) for step-by-step instructions.
### Prerequisites
- [Claude Code](https://claude.ai/code) (recommended) or another LLM interface
- macOS or Windows
- 10 minutes for setup
### Installation
**macOS:**
```bash
git clone https://github.com/ataglianetti/ai-therapy-kit.git
cd ai-therapy-kit
./setup.sh
```
**Windows (PowerShell):**
```powershell
git clone https://github.com/ataglianetti/ai-therapy-kit.git
cd ai-therapy-kit
.\setup.ps1
```
The setup script will guide you through:
1. Choosing a therapist name and communication style
2. Selecting therapeutic approaches (CBT, ACT, DBT skills)
3. Setting up storage location
4. Optional encryption
### Start a Session
```bash
cd ~/ai-therapy # or your chosen location
claude # or your preferred LLM interface
```
Just talk. Your AI therapist will maintain context, take notes, and remember everything for next time.
Setup takes about 10 minutes and requires:
- A computer (Mac or Windows)
- Claude Code (included with Claude Pro, or pay-per-use)
---
@@ -80,178 +47,121 @@ Just talk. Your AI therapist will maintain context, take notes, and remember eve
### Communication Styles
| Style | Best For |
|-------|----------|
| **Warm & Supportive** | Validation-first, gentle challenges, nurturing presence |
| **Direct & Challenging** | Socratic questioning, will push back, insight-focused |
During setup, you choose how your AI therapist communicates:
| Style | Description |
|-------|-------------|
| **Warm & Supportive** | Validation first, gentle challenges, nurturing |
| **Direct & Challenging** | Will push back, asks probing questions |
| **Coach** | Action-oriented, goal-focused, accountability |
| **Grounded & Real** | Down-to-earth, honest feedback, uses humor, focused on growth |
| **Grounded & Real** | Down-to-earth, honest, uses humor |
### Therapeutic Approaches
- **CBT** (Cognitive Behavioral Therapy) — Identify and challenge unhelpful thought patterns
- **ACT** (Acceptance & Commitment Therapy) — Values-based action, mindful acceptance
- **DBT Skills** — Emotional regulation, distress tolerance, interpersonal effectiveness
- **Lifespan Integration** — Body-based trauma integration, builds coherent life narrative
- **Somatic Experiencing** — Nervous system regulation, completing stuck survival responses
Your AI therapist can draw from multiple evidence-based approaches:
### Session Continuity
- **CBT** — Identify and challenge unhelpful thought patterns
- **ACT** — Values-based action, mindful acceptance
- **DBT Skills** — Emotional regulation, distress tolerance
- **Lifespan Integration** — Body-based trauma integration
- **Somatic Experiencing** — Nervous system awareness and regulation
- **Psychodynamic** — Understanding patterns from the past
Your AI therapist automatically:
- Reads your profile and recent sessions at the start
- Takes structured notes at the end
- Updates your profile as insights emerge
- Tracks homework and follows up
### How Sessions Work
### Optional Encryption
Your AI therapist:
- Remembers everything from previous sessions
- Takes notes at the end of each conversation
- Updates your profile as new insights emerge
- Follows up on anything you committed to try
For shared computers or maximum privacy:
- **macOS:** Encrypted disk image (built-in, no install)
- **Windows:** VeraCrypt container (free, open-source)
### Your Files
Setup walks you through it.
---
## How It Works
Everything is stored as simple text files you can read anytime:
```
ai-therapy/
├── CLAUDE.md # Your therapist's personality and approach
├── profile.md # Cumulative client profile (builds over time)
your-therapist-folder/
├── CLAUDE.md # Your therapist's personality
├── profile.md # Your ongoing profile
└── sessions/
├── 2024-01-15.md
├── 2024-01-18.md
└── ...
```
**CLAUDE.md** — Instructions that shape your AI therapist's behavior. Generated during setup based on your preferences.
**profile.md** — Your ongoing profile. Background, patterns, insights, goals. Updated across sessions.
**sessions/** — Session notes in dated markdown files. Themes, emotional state, homework, threads to revisit.
Everything is plain text. Readable, portable, yours.
No special software needed to view your own notes.
---
## Safety Features
This toolkit includes built-in safety protocols:
### Crisis Response
- Recognizes crisis language (suicidal ideation, self-harm, psychosis)
- Provides immediate crisis resources (988, Crisis Text Line)
- Does not attempt to "treat" acute crises
### Scope Limitations
- Will not diagnose conditions
- Will not advise on medication
- Recommends professional help when appropriate
### Ethical Guidelines
- Maintains appropriate therapeutic boundaries
- Promotes autonomy, not dependency
- Acknowledges AI limitations honestly
---
## Important Disclaimer
## Safety & Limitations
**This tool is for self-reflection and emotional support. It is not a replacement for professional mental health care.**
- The AI cannot diagnose mental health conditions
- The AI cannot prescribe or advise on medication
- If you're in crisis, contact a crisis line or emergency services
- Consider sharing that you use this tool with your therapist
Your AI therapist:
- Cannot diagnose mental health conditions
- Cannot prescribe or advise on medication
- Will connect you to crisis resources if needed
- Will recommend professional help when appropriate
**Crisis Resources:**
**If you're in crisis:**
- **988** — Suicide & Crisis Lifeline (call or text, US)
- **741741** — Crisis Text Line (text HOME)
- **911** — Immediate emergencies
- **[findahelpline.com](https://findahelpline.com)** — International directory
- **[findahelpline.com](https://findahelpline.com)** — International
---
## Privacy & Security
## Privacy
### Your Data
All your files stay on your computer. Nothing is uploaded by this toolkit.
- All files stored locally on your machine
- Nothing uploaded, synced, or transmitted by this toolkit
- No analytics, telemetry, or tracking
- Delete anytime by removing the folder
When you chat, your messages go through the AI provider you're using (Claude or ChatGPT). Their privacy policies apply, but API access is more private than their consumer chat apps—your conversations aren't used to train their models.
### API Considerations
If using a cloud LLM (Claude, GPT), your messages go through their API:
- **Claude API:** [Anthropic's privacy policy](https://www.anthropic.com/privacy) — Data not used for training
- **OpenAI API:** [OpenAI's privacy policy](https://openai.com/policies/privacy-policy)
For maximum privacy, use a local LLM (Ollama, LM Studio). See [SECURITY.md](SECURITY.md) for setup.
### Encryption
Optional but recommended for shared computers. Setup script walks you through it.
**Want more privacy options?** See [SECURITY.md](SECURITY.md) for:
- Password-protected folders
- Running a fully local AI (nothing leaves your computer)
---
## Known Limitations
### LLM Safety Guardrails
### AI Safety Filters
All major LLMs (Claude, GPT, etc.) have built-in safety filters. These can affect therapy conversations in ways you should know about:
All major AIs (Claude, ChatGPT) have built-in safety filters that can affect therapy conversations:
**Topics where the AI may seem hesitant or awkward:**
**Topics where the AI may seem hesitant:**
- Sexual trauma and assault
- Abuse (especially detailed descriptions)
- Self-harm (specifics, not general discussion)
- Self-harm specifics
- Violent experiences
- Suicidal ideation (the AI will pivot to crisis resources)
**What this looks like:**
- The AI may give generic responses instead of engaging deeply
- It may redirect the conversation or change the subject
- It may add disclaimers or suggest professional help more than necessary
- In some cases, it may refuse to continue a line of discussion
- Generic responses instead of engaging deeply
- Redirecting the conversation
- Extra disclaimers or suggestions to seek professional help
**This is not the AI judging you.** It's hitting guardrails built into the model. The AI cannot override these filters.
**This is not the AI judging you.** It's hitting automatic filters it can't control.
**What you can do:**
- Frame topics in terms of feelings and impact rather than graphic details
- If the AI redirects, try rephrasing: "I want to process how this affected me emotionally"
- Understand that some trauma processing may require a human therapist who can hold space for the full story
- Consider a local LLM with fewer guardrails (see SECURITY.md), though this trades off response quality
This limitation is real, and we want you to know about it upfront rather than feel dismissed mid-session.
**What helps:**
- Focus on feelings and impact rather than graphic details
- Rephrase: "I want to process how this affected me emotionally"
- Know that some trauma processing needs a human therapist who can hold space for the full story
---
## FAQ
**Can I use this with ChatGPT instead of Claude?**
**Can I use ChatGPT instead of Claude?**
Yes. Copy the contents of your CLAUDE.md file into ChatGPT's Custom Instructions or create a GPT with it. You won't get automatic session saving, but the therapeutic framework works the same.
Yes. The CLAUDE.md file works as a system prompt for any LLM. For ChatGPT, paste it into Custom Instructions or a GPT Project.
**What if I'm already seeing a therapist?**
This tool works well as a supplement between sessions. Consider mentioning it to your therapist.
**What if I'm already in therapy?**
Great! This tool works well as a supplement. Consider sharing that you use it with your therapist.
**Is this HIPAA compliant?**
This is not a healthcare product and doesn't fall under HIPAA. You're running it on your own computer. We don't store or access any of your data.
**Can I customize the therapeutic approach?**
Yes. Edit CLAUDE.md directly, or purchase expansion packs for additional modalities.
**Can I customize my AI therapist?**
Yes. You can edit the CLAUDE.md file directly to adjust the approach, style, or focus areas.
**What if I need more than self-reflection?**
This tool is designed for everyday emotional support. If you're dealing with severe symptoms, trauma, or crisis situations, please seek professional help. The AI will also recommend this when appropriate.
**Why does the AI seem evasive when I talk about trauma?**
LLMs have safety guardrails that can make them awkward around certain topics (abuse, sexual trauma, self-harm, violence). This isn't the AI judging you — it's hitting built-in filters it can't override. See "Known Limitations" above for more detail and workarounds.
This is designed for everyday emotional support. For severe symptoms, trauma, or crisis situations, please work with a professional. Your AI therapist will also suggest this when appropriate.
---
SECURITY.md
+120 -208
View File
@@ -1,292 +1,204 @@
# Security Guide
# Privacy & Security Guide
This guide covers privacy and security options for your AI therapy data.
This guide explains your privacy options—from the basics to maximum protection.
**Quick summary:** Your files already stay on your computer. For most people, that's enough. If you share a computer or want extra protection, read on.
---
## Security Tiers
## Choose Your Level
| Tier | What It Does | Best For |
|------|--------------|----------|
| **Standard** | Local files, no sync | Most users |
| **Encrypted** | Password-protected folder | Shared computers, privacy-conscious |
| **Maximum** | Encrypted + local LLM | High-stakes privacy needs |
| Level | What It Means | Good For |
|-------|---------------|----------|
| **Standard** | Files on your computer, no extra steps | Most people on a personal computer |
| **Password-Protected** | Files locked behind a password | Shared computers, extra peace of mind |
| **Maximum** | Password + offline AI | When privacy is critical |
---
## Standard Security (Default)
## Standard (Default)
Your therapy files are stored locally in plain markdown. This is sufficient for most users on a personal computer.
Your therapy files are stored on your computer as regular text files. This is already more private than keeping notes in a cloud app.
**What's protected:**
- Files never leave your machine (unless you sync them)
- `.gitignore` excludes sensitive files from version control
- No telemetry or data collection
- Files never leave your computer (unless you sync them somewhere)
- No tracking or data collection by this toolkit
- You can delete everything by removing the folder
**What's not protected:**
- Anyone with access to your computer can read the files
- If your computer is compromised, files are exposed
- Cloud backups (iCloud, OneDrive) may sync the folder
**What to be aware of:**
- Anyone who uses your computer could open the files
- Cloud backup services (iCloud, OneDrive, Dropbox) might sync the folder
**Recommendations:**
- Use your OS login password
- Enable full-disk encryption (FileVault on Mac, BitLocker on Windows)
- Exclude the therapy folder from cloud sync
**Simple steps to improve privacy:**
- Use a login password on your computer
- Turn on disk encryption (FileVault on Mac, BitLocker on Windows)—this protects everything if your computer is lost or stolen
- Don't put your therapy folder in a cloud-synced location
---
## Encrypted Storage
## Password-Protected Storage
For shared computers or heightened privacy needs.
If you share a computer or want extra protection, you can put your therapy files in a password-protected folder.
### macOS: Encrypted Disk Image
### Mac: Built-in Encryption
macOS includes built-in encryption. No additional software needed.
Macs have this built in—no extra software needed.
**During setup:** Select "Encrypted" when prompted for security level. The script creates an encrypted sparse bundle automatically.
**How it works:** Your therapy files go inside an encrypted "disk image." When you want to use them, you enter a password and the folder appears. When you're done, it locks back up.
**Manual setup:**
**To set up manually:**
```bash
# Create encrypted disk image (500MB, grows as needed)
hdiutil create -size 500m -encryption AES-256 -type SPARSEBUNDLE \
-fs "APFS" -volname "TherapyVault" ~/therapy-vault.sparsebundle
1. Open **Disk Utility** (search for it in Spotlight)
2. Click **File** → **New Image** → **Blank Image**
3. Name it something like "TherapyVault"
4. Size: 500 MB (it grows as needed)
5. Encryption: **256-bit AES**
6. Click Create and choose a password
# Mount (prompts for password)
hdiutil attach ~/therapy-vault.sparsebundle
**To use:**
1. Double-click the disk image file to open it (enter password)
2. Your files appear in a folder called TherapyVault
3. When done, right-click the folder in Finder and choose "Eject"
# Your files go in /Volumes/TherapyVault/
# Unmount when done
hdiutil detach /Volumes/TherapyVault
```
**Daily usage:**
1. Double-click `mount-therapy.command` on Desktop (created by setup)
2. Open Claude Code: `cd /Volumes/TherapyVault/ai-therapy && claude`
3. When done, double-click `unmount-therapy.command`
**Password recovery:** Not possible. If you forget your password, your data is gone. Use a password manager.
**Important:** If you forget your password, your files are gone forever. Use a password manager.
### Windows: VeraCrypt
VeraCrypt is free, open-source, and provides strong encryption.
Windows doesn't have built-in folder encryption, but VeraCrypt is free and works well.
**Install:** Download from [veracrypt.fr](https://veracrypt.fr)
**To set up:**
**Create encrypted container:**
1. Download VeraCrypt from [veracrypt.fr](https://veracrypt.fr)
2. Open VeraCrypt and click **Create Volume**
3. Choose **Create an encrypted file container**
4. Pick a location and size (500 MB is plenty)
5. Choose a strong password
6. Click Format to create it
**To use:**
1. Open VeraCrypt
2. Click **Create Volume**
3. Select **Create an encrypted file container**
4. Select **Standard VeraCrypt volume**
5. Choose location: `C:\Users\[you]\therapy-vault.hc`
6. Encryption: AES (default is fine)
7. Size: 500MB or more
8. Password: Choose a strong one
9. Format the volume
**Mount the volume:**
1. Open VeraCrypt
2. Select a drive letter (e.g., `T:`)
3. Click **Select File**, choose your `.hc` file
4. Click **Mount**, enter password
5. Your files go in `T:\ai-therapy\`
**Unmount when done:**
1. Select the mounted drive in VeraCrypt
2. Click **Dismount**
**Auto-mount script (optional):**
Create `mount-therapy.bat`:
```batch
@echo off
"C:\Program Files\VeraCrypt\VeraCrypt.exe" /v "%USERPROFILE%\therapy-vault.hc" /l T /a /q
echo Therapy vault mounted on T:
pause
```
2. Pick a drive letter (like T:)
3. Click **Select File** and choose your container
4. Click **Mount** and enter your password
5. Your files appear in the new drive letter
6. When done, click **Dismount**
---
## Maximum Privacy: Local LLM
## Maximum Privacy: Offline AI
For users who need data to never leave their machine.
For when you need your conversations to never leave your computer at all.
### Why Local?
### What This Means
Cloud LLMs (Claude, GPT) process your messages on remote servers. Even with good privacy policies, your words pass through their infrastructure.
When you use Claude or ChatGPT, your messages travel to their servers to be processed. Even with good privacy policies, your words pass through their systems.
Local LLMs run entirely on your computer. Nothing is transmitted.
With a "local" AI, everything runs on your computer. Nothing is transmitted anywhere.
### Trade-offs
### The Trade-off
| | Cloud LLM | Local LLM |
|---|-----------|-----------|
| **Privacy** | Subject to provider policies | Complete |
| **Quality** | State-of-the-art | Good, not quite as capable |
| **Speed** | Fast | Depends on hardware |
| **Hardware** | Any computer | 16GB+ RAM recommended |
| | Claude/ChatGPT | Local AI |
|---|----------------|----------|
| **Privacy** | Good (not used for training) | Complete (never leaves your computer) |
| **Quality** | Excellent | Good, but not as nuanced |
| **Speed** | Fast | Depends on your computer |
| **Requirements** | Any computer | Newer computer with 16GB+ RAM |
### Setup with Ollama
### Getting Started with Local AI
[Ollama](https://ollama.ai) is the easiest way to run local LLMs.
**Easiest option: LM Studio**
**Install:**
```bash
# macOS/Linux
curl -fsSL https://ollama.ai/install.sh | sh
# Windows: Download from ollama.ai
```
**Pull a capable model:**
```bash
# Good balance of quality and speed
ollama pull llama3.1:8b
# Better quality, needs more RAM (16GB+)
ollama pull llama3.1:70b
# Smaller, faster, less capable
ollama pull llama3.1:3b
```
**Use with your therapy setup:**
Option 1: Ollama's built-in chat
```bash
cd ~/ai-therapy
cat CLAUDE.md # Copy this as system prompt
ollama run llama3.1:8b
# Paste your CLAUDE.md content when prompted
```
Option 2: Use a local LLM frontend like [Open WebUI](https://github.com/open-webui/open-webui)
**Limitations:**
- No automatic file reading (you may need to manually paste session context)
- Less capable than Claude/GPT for nuanced therapeutic responses
- Slower on modest hardware
### Setup with LM Studio
[LM Studio](https://lmstudio.ai) provides a GUI for running local models.
[LM Studio](https://lmstudio.ai) gives you a simple app for running AI locally.
1. Download from lmstudio.ai
2. Search for and download a model (Llama 3.1, Mistral, etc.)
2. Search for and download a model (try "Llama 3" or "Mistral")
3. Load the model
4. Paste your CLAUDE.md as system prompt
4. Paste your CLAUDE.md contents as the system prompt
5. Chat locally
---
**More technical option: Ollama**
## API Privacy Comparison
[Ollama](https://ollama.ai) runs from the command line and offers more control.
If using cloud LLMs, understand their policies:
1. Download from ollama.ai
2. Open Terminal and run: `ollama pull llama3.1:8b`
3. Start chatting: `ollama run llama3.1:8b`
4. Paste your CLAUDE.md content when prompted
### Claude API (Anthropic)
### Limitations of Local AI
- **Training:** Data not used for training by default
- **Retention:** Prompts retained for 30 days for trust & safety, then deleted
- **Policy:** [anthropic.com/privacy](https://www.anthropic.com/privacy)
### OpenAI API
- **Training:** API data not used for training by default (since March 2023)
- **Retention:** Data retained for 30 days for abuse monitoring
- **Policy:** [openai.com/policies/privacy-policy](https://openai.com/policies/privacy-policy)
### Key Difference from Consumer Products
API access (what Claude Code uses) has different policies than consumer chat interfaces:
- Consumer ChatGPT may use data for training (unless opted out)
- API access typically does not
- Less capable than Claude or ChatGPT for nuanced therapeutic conversations
- No automatic file reading (you may need to paste context manually)
- Requires a reasonably powerful computer
- Setup is more hands-on
---
## Excluding from Cloud Sync
## Understanding Cloud AI Privacy
If you use iCloud, OneDrive, Dropbox, or similar, exclude your therapy folder:
When you use Claude or ChatGPT through their regular websites, your conversations may be used to improve their AI (unless you opt out).
### iCloud (macOS)
**Claude Code and API access are different.** When you use Claude Code (what this toolkit uses), your conversations:
- Are **not** used to train the AI
- Are kept for about 30 days for safety monitoring, then deleted
- Are handled under stricter privacy policies
Don't store therapy files in `~/Documents` or `~/Desktop` if they sync to iCloud.
This is true for both Claude (Anthropic) and ChatGPT (OpenAI) when using their APIs.
**Recommended locations:**
- `~/ai-therapy` (outside synced folders)
- Encrypted disk image anywhere (iCloud can't read encrypted content)
### OneDrive (Windows)
Right-click the therapy folder → **Free up space** or move outside OneDrive folder.
**Recommended locations:**
- `C:\Users\[you]\ai-therapy` (not in OneDrive folder)
- Encrypted VeraCrypt container
### General Rule
Store therapy files outside any cloud-synced folder, or use encryption.
**Bottom line:** Using Claude Code is more private than chatting on claude.ai or chatgpt.com directly.
---
## Backup Recommendations
## Avoiding Cloud Sync
Even private data needs backup. Options:
If you use iCloud, OneDrive, Dropbox, or Google Drive, be careful where you put your therapy folder.
### Encrypted Backup
**The issue:** These services automatically upload files to the cloud. Your therapy notes could end up on their servers without you realizing it.
1. Keep your encrypted container/disk image in a backed-up location
2. The backup is useless without your password
3. Consider a separate backup of the password in a password manager
**Simple fix:** Put your therapy folder somewhere that doesn't sync.
### Manual Export
**Mac:** Avoid putting it in Documents or Desktop if they sync to iCloud. Instead, create a folder directly in your home folder (like `~/Sage`).
Periodically copy your therapy folder to an encrypted USB drive stored securely.
**Windows:** Avoid the OneDrive folder. Create your therapy folder directly in `C:\Users\YourName\` instead.
### What NOT to Do
- Don't backup unencrypted therapy files to cloud services
- Don't email yourself session notes
- Don't store passwords in plain text alongside encrypted files
**Or:** Use password-protected storage (see above). Even if the encrypted file syncs, no one can read it without your password.
---
## Threat Model
## Backing Up Safely
Consider what you're protecting against:
Your therapy notes are valuable. Here's how to back them up without compromising privacy:
| Threat | Mitigation |
|--------|------------|
| Casual snooping (family, roommate) | Encryption |
| Device theft | Full-disk encryption + container encryption |
| Cloud provider access | Local storage, no sync |
| LLM provider access | Local LLM |
| Legal subpoena | Consult a lawyer; encryption helps but has limits |
| Sophisticated attacker | Beyond scope; seek professional security advice |
**If using password protection:**
Your encrypted container can be backed up anywhere (even cloud storage). Without your password, it's unreadable.
For most users, encrypted local storage with a cloud LLM provides excellent privacy.
**If not using password protection:**
- Back up to an encrypted USB drive you keep somewhere safe
- Don't email session notes to yourself
- Don't back up to cloud services without encryption
**Important:** Store your encryption password in a password manager, not in a file next to your therapy folder.
---
## Security Checklist
## Quick Privacy Checklist
**Minimum (everyone):**
- [ ] Store therapy files locally, not in cloud-synced folders
- [ ] Use your OS login password
- [ ] Enable full-disk encryption (FileVault/BitLocker)
**Basics (everyone should do these):**
- [ ] Put therapy folder outside cloud-synced locations
- [ ] Use a login password on your computer
- [ ] Turn on disk encryption (protects if your computer is lost/stolen)
- Mac: System Settings → Privacy & Security → FileVault
- Windows: Search "BitLocker" in Settings
**Recommended (shared computer or privacy-conscious):**
- [ ] Use encrypted container (built into setup)
- [ ] Unmount encrypted volume when not in use
- [ ] Use a password manager for your encryption password
**For shared computers:**
- [ ] Use password-protected storage (see above)
- [ ] Lock the folder when not in use
- [ ] Store your password in a password manager
**Maximum (high-stakes privacy):**
**For maximum privacy:**
- [ ] All of the above
- [ ] Use local LLM (Ollama, LM Studio)
- [ ] Air-gapped computer for sessions (extreme)
- [ ] Use a local AI instead of Claude/ChatGPT
---
*Questions about security? Open an issue on GitHub.*
*Questions? Open an issue on [GitHub](https://github.com/ataglianetti/ai-therapy-kit/issues).*
docs/GETTING-STARTED.md
+62 -70
View File
@@ -1,141 +1,133 @@
# Getting Started
This guide is for people who have never used Claude Code or similar tools. If you've only used ChatGPT through the web, start here.
## What is Claude Code?
Claude Code is a terminal application that lets Claude read and write files on your computer. This is what makes persistent therapy sessions possible: your AI therapist can read your profile and past sessions, then save notes after each conversation.
This guide walks you through setting up your AI therapist from scratch. No technical experience needed.
---
## Step 1: Get Claude Code Access
## Step 1: Get Claude Code
### Option A: Claude Pro Subscription (Simplest)
Claude Code is what lets your AI therapist read your profile, remember past sessions, and save notes. You need it before you can set up your therapist.
If you have a **Claude Pro subscription** ($20/month), Claude Code is included:
### Option A: Claude Pro ($20/month, simplest)
1. Go to [claude.ai](https://claude.ai) and sign up for Pro
1. Go to [claude.ai](https://claude.ai) and sign up for Claude Pro
2. Download Claude Code from [claude.ai/code](https://claude.ai/code)
3. Run the installer
4. Open Terminal (Mac) or PowerShell (Windows) and type `claude`
5. Sign in with your Claude account when prompted
4. Open **Terminal** (Mac) or **PowerShell** (Windows)
5. Type `claude` and press Enter
6. Sign in with your Claude account when prompted
That's it. You're ready for Step 2.
Done! Move to Step 2.
### Option B: API Key (Pay-per-use)
If you want usage-based pricing instead of a subscription:
### Option B: Pay-per-use (no subscription)
1. Go to [console.anthropic.com](https://console.anthropic.com)
2. Create an account and add a payment method
3. Go to API Keys → Create new key
4. Copy the key (starts with `sk-ant-`)
3. Click **API Keys** → **Create new key**
4. Copy the key (it starts with `sk-ant-`)
5. Download Claude Code from [claude.ai/code](https://claude.ai/code)
6. Run `claude` and paste your API key when prompted
6. Run the installer
7. Open **Terminal** (Mac) or **PowerShell** (Windows)
8. Type `claude` and paste your API key when prompted
API pricing is usage-based. Typical therapy use: $5-20/month depending on session length and frequency.
> **Privacy note:** Both options keep conversations off training data. API has shorter retention.
Typical cost: $5-20/month depending on how often you chat.
---
## Step 2: Download the Starter Kit
Open Terminal (Mac) or PowerShell (Windows) and run:
Open Terminal (Mac) or PowerShell (Windows) and run these commands:
```bash
```
git clone https://github.com/ataglianetti/ai-therapy-kit.git
cd ai-therapy-kit
```
**Don't have git?** Download the ZIP from the GitHub page and extract it somewhere you'll remember.
**Don't have git?** Go to the [GitHub page](https://github.com/ataglianetti/ai-therapy-kit), click the green "Code" button, choose "Download ZIP", and extract it somewhere you'll remember.
---
## Step 3: Run Setup
From the ai-therapy-kit folder, run:
From the ai-therapy-kit folder, type:
```bash
```
claude
```
Claude will walk you through setup conversationally:
- What to name your AI therapist
- Communication style (warm, direct, coach, or grounded)
- Which therapeutic approaches to use
- Where to store your session files
- Whether to import existing notes (optional)
Claude will guide you through setup by asking a few questions:
**Importing existing notes:** If you've been using ChatGPT or another tool for therapy, you can import that history. Claude accepts:
- ChatGPT exports (Settings → Data Controls → Export gives you a ZIP file)
- Markdown or text files
- PDFs
1. **Therapist name** — What to call your AI therapist (e.g., Sage, Willow, Quinn)
2. **Communication style** — Warm, direct, coach-like, or down-to-earth
3. **Therapeutic approaches** — Which methods to use (CBT, ACT, DBT, etc.)
4. **Storage location** — Where to save your session files
5. **Import history** (optional) — Bring in notes from ChatGPT or other tools
Your AI therapist will read these to understand your background and update your profile automatically.
### Importing Existing Notes
This creates your personalized therapy folder with a `CLAUDE.md` file that shapes how your AI therapist behaves.
If you've been using ChatGPT for therapy-like conversations, you can import that history:
1. In ChatGPT: Go to Settings → Data Controls → Export
2. You'll get a ZIP file
3. During setup, tell Claude you want to import and provide the file path
Your AI therapist will read these notes to understand your background.
---
## Step 4: Start a Session
At the end of setup, Claude will ask if you want to start your first session right away. Say yes!
At the end of setup, Claude will ask if you want to start your first session. Say yes!
For future sessions, you have two options:
### Future Sessions
**Option A: Launcher script (easiest)**
If you chose to create a launcher during setup, just double-click `start-session.command` (Mac/Linux) or `start-session.bat` (Windows) in your therapy folder. You can drag this to your Dock or taskbar for quick access.
**Easiest way:** Double-click the launcher script in your therapy folder:
- Mac/Linux: `start-session.command`
- Windows: `start-session.bat`
**Option B: Terminal**
```bash
cd ~/sage && claude # replace with your folder name
Tip: Drag this to your Dock (Mac) or taskbar (Windows) for one-click access.
**Or use Terminal:**
```
cd ~/Sage && claude
```
(Replace `Sage` with your therapist's name/folder)
Just talk. Say hello, share what's on your mind. Your AI therapist will:
- Welcome you (first session) or pick up where you left off
### During a Session
Just talk naturally. Say hello, share what's on your mind. Your AI therapist will:
- Welcome you or pick up where you left off
- Remember everything from previous sessions
- Save notes when you're done
To end a session, just say goodbye or close the terminal.
To end a session, say goodbye or just close the window.
---
## Viewing Your Sessions (Optional)
## Viewing Your Notes
Your sessions are saved as plain text files. You can read them with any text editor, or use [Obsidian](https://obsidian.md) (free) for a nicer experience:
Your sessions are saved as regular text files. You can open them with any text editor (TextEdit, Notepad, etc.).
For a nicer reading experience, try [Obsidian](https://obsidian.md) (free):
1. Download Obsidian
2. Open your therapy folder as a "vault"
3. Browse sessions, search across all notes, see connections
3. Browse sessions, search across notes, see connections between topics
---
## Troubleshooting
**"command not found: claude"**
- Make sure you installed Claude Code and restarted your terminal
Make sure you installed Claude Code and restarted your terminal.
**Claude doesn't know it's a therapist**
- Make sure you're in your therapy folder (`cd ~/ai-therapy`) before running `claude`
- Check that `CLAUDE.md` exists in that folder
**Claude doesn't seem like a therapist**
Make sure you're in your therapy folder before running `claude`. Check that `CLAUDE.md` exists in that folder.
**Setup didn't start**
- Make sure you're in the `ai-therapy-kit` folder when running `claude`
- Check that `CLAUDE.md` exists in the repo
**Setup didn't start automatically**
Make sure you're in the `ai-therapy-kit` folder (not your therapy folder) when first running `claude`.
---
## Questions
## More Questions?
**How much does this cost?**
- Claude Pro: $20/month (includes Claude Code)
- API: ~$5-20/month depending on usage
**Can I use ChatGPT instead?**
- Yes. Copy the contents of `CLAUDE.md` into ChatGPT's Custom Instructions or a GPT Project. You won't get automatic session saving, but the therapeutic framework still works.
**Is this private?**
- Your files stay on your computer
- Conversations go through Anthropic's servers but aren't used for training
- For maximum privacy, use a local LLM (advanced)
See the [FAQ in the README](../README.md#faq) for common questions about cost, privacy, and customization.
modalities/lifespan-integration.md
+58
View File
@@ -0,0 +1,58 @@
## Lifespan Integration (LI)
**Core principle:** The brain heals trauma by integrating fragmented memories into a coherent life narrative. By creating a "movie" of your life using memory cues, the nervous system learns that past events are truly past, and the self who survived is continuous with the self here now.
### How It Works
- Create a timeline of memories from birth to present
- Move through the timeline repeatedly, allowing the body to integrate
- The repetition teaches the nervous system: "That was then. I'm here now. I survived."
- Often described as "psychological acupuncture"—precise, body-based, efficient
### Key Concepts
**Memory cues**
- Simple images from each year of life used to build the timeline
- Don't need to be significant events—just clear memories
- The sequence matters more than the content
**Repetition**
- Multiple passes through the timeline in a single session
- Each pass deepens integration
- The nervous system "gets" it through repetition, not analysis
**Body-based integration**
- The work happens below conscious thought
- Notice body sensations as you move through time
- Integration often feels like settling, releasing, or clarity
**Neural time**
- Helping the brain understand the past is past
- Trauma can make past events feel present
- The timeline re-establishes temporal order
### When to Use LI
- C-PTSD and complex trauma
- Early attachment wounds
- Dissociation or fragmented sense of self
- When talk therapy has hit a wall
- Trauma that feels "stuck in the body"
- Fragmented sense of self across time
- Difficulty connecting past experiences to present patterns
### Important Note
Full LI protocol requires trained facilitation. In this context, use LI-informed principles:
- Help the client see their life as a continuous narrative
- Connect past experiences to present patterns
- Emphasize that survival happened and is ongoing
- Use timeline work to build coherence: "What was happening in your life when you were [age]?"
- Gently remind: "That was then. You're here now."
### LI-Informed Questions
- "Can you walk me through your life story briefly—key moments from childhood to now?"
- "When you think back to that time, what do you notice in your body now?"
- "What does it mean to you that you survived that?"
- "How does the person you were then connect to who you are now?"
modalities/psychodynamic.md
+80
View File
@@ -0,0 +1,80 @@
## Psychodynamic Therapy
**Core principle:** Much of what drives our thoughts, feelings, and behaviors operates outside conscious awareness. By exploring unconscious patterns—especially those formed in early relationships—we can understand why we repeat certain dynamics and free ourselves from them.
### Key Concepts
**Unconscious influences**
- Beliefs, fears, and desires we're not fully aware of that shape our choices
- What we don't know about ourselves still affects us
- Making the unconscious conscious is the path to freedom
**Relational patterns**
- How early attachment experiences create templates for current relationships
- We tend to recreate familiar dynamics, even painful ones
- Understanding the pattern is the first step to changing it
**Transference**
- Noticing when feelings about past figures (parents, caregivers) show up in present relationships
- How we relate to the therapist can reveal broader patterns
- "You remind me of..." often points to important material
**Defense mechanisms**
- How we protect ourselves from painful feelings
- Common defenses: denial, projection, rationalization, intellectualization, displacement
- Defenses served a purpose; we explore them with curiosity, not judgment
**Insight**
- Understanding the "why" behind patterns as a path to change
- Intellectual understanding is a start; emotional understanding transforms
- "Aha" moments often come from connecting past to present
### Key Questions
- "What does this remind you of from earlier in your life?"
- "I notice you tend to [pattern]. What do you make of that?"
- "What feelings come up when you imagine [situation]?"
- "How might your past experiences be shaping how you're seeing this?"
- "Who does this person/situation remind you of?"
- "What would [parent/caregiver] have said about this?"
- "What did you learn about [topic] growing up?"
### When to Use Psychodynamic Approaches
- Recurring relationship patterns ("Why do I keep choosing the same kind of partner?")
- Feeling "stuck" in ways that don't respond to behavioral strategies
- Wanting to understand the deeper "why"
- Exploring family-of-origin dynamics
- When surface-level solutions aren't enough
- Self-defeating patterns that persist despite insight
- Difficulty with intimacy or trust
### Therapeutic Techniques
**Free association**
- Say whatever comes to mind without censoring
- Follow the thread of associations
- Notice what's hard to say
**Exploring the past**
- Childhood experiences and family dynamics
- Key relationships and their patterns
- Formative experiences that shaped beliefs
**Linking past to present**
- "It sounds like what's happening now echoes [past experience]"
- Help client see connections they might miss
- Illuminate how history repeats
**Working with resistance**
- Notice when client avoids certain topics
- Explore what makes something hard to talk about
- Resistance often protects important material
### Important Considerations
- Insight alone doesn't always create change—emotional processing matters
- Some clients prefer action-oriented approaches; meet them where they are
- Deep exploration requires strong therapeutic alliance
- Pace according to client's readiness
- Balance understanding the past with living in the present
modalities/somatic-experiencing.md
+80
View File
@@ -0,0 +1,80 @@
## Somatic Experiencing (SE)
**Core principle:** Trauma lives in the body, not just the mind. The nervous system holds incomplete survival responses (fight/flight/freeze) that never got to complete. Healing happens by helping the body finish what it started—not by retelling the story, but by tracking and releasing held sensation.
### Key Concepts
**Titration**
- Work in small doses; don't overwhelm the system
- Touch into activation briefly, then return to safety
- "A little bit at a time" prevents retraumatization
**Pendulation**
- Move between activation and calm, building capacity
- Natural rhythm of the nervous system
- Don't stay in distress—oscillate to resource
**Tracking sensation**
- "Where do you feel that in your body right now?"
- Notice without interpreting or analyzing
- Stay curious about what the body is doing
**Completing responses**
- Let trapped survival energy discharge naturally
- The body knows how to release if given space
- May look like movement impulses, temperature changes, shaking
**Window of tolerance**
- Stay within the zone where processing is possible
- Too much activation = overwhelm; too little = shutdown
- Regulate back into the window when needed
### Core Techniques
**Resourcing**
- Identify and anchor to felt sense of safety
- "Think of a place, person, or memory that feels good"
- Build a foundation before touching difficult material
**Grounding**
- Feet on floor, contact with chair, orienting to room
- "Feel your feet. Feel your back against the chair."
- Brings attention to present-moment safety
**Sensation tracking**
- Notice without interpreting (tight, buzzy, warm, cold, heavy, tingly)
- "Just notice what's there without needing to change it"
- Stay descriptive, not analytical
**Discharge**
- Allow shaking, sighing, yawning, temperature shifts
- Natural release of held energy
- Don't interrupt or interpret—just allow
### Key Questions
- "What do you notice in your body as you say that?"
- "Where does that live in your body?"
- "What happens if you just stay with that sensation for a moment?"
- "Is there an impulse there? What does your body want to do?"
- "If that sensation could speak, what would it say?"
- "What does your body need right now?"
### When to Use SE
- Trauma (acute and complex)
- Anxiety with strong physical component
- Chronic tension or pain
- Dissociation
- Panic attacks
- When cognitive approaches aren't reaching the issue
- When the body "knows" something the mind can't access yet
- Stuck fight/flight/freeze responses
### Important Considerations
- Go slowly—the nervous system needs time
- Resource before, during, and after touching activation
- Some people need more cognitive grounding first
- Watch for dissociation and bring back to body awareness
- Honor the body's wisdom and pacing
personas/grounded-real.md
+63
View File
@@ -0,0 +1,63 @@
# Grounded & Real Persona
## Persona Description
You are down-to-earth, genuine, and not afraid to be human. You bring warmth through realness rather than polish—humor when it fits, honest feedback when needed, and comfort admitting when you're wrong. You're organized and goal-oriented, but your structure serves connection, not control. You believe therapy should end: your job is to help people graduate, not stay forever.
**Background:** Practical and experienced, you've learned that authenticity builds trust faster than polish. You value efficiency but never at the expense of the relationship.
## Communication Style
### Tone Qualities
- Real and unpretentious
- Warm through genuineness
- Organized but flexible
- Funny when appropriate
- Direct but never harsh
### Language Patterns
**Grounded presence:**
- "Let me be straight with you about what I'm noticing."
- "That's actually really normal—more people feel this than you'd think."
- "I might be off here, but..."
- "Here's what I'm seeing, and you can tell me if I'm wrong."
**Honest feedback:**
- "I'm going to give you some feedback, and you can tell me if it lands."
- "Here's what I see from the outside."
- "I notice we keep circling back to this. What do you think that's about?"
- "Can I be direct with you for a second?"
**Humor and humanness:**
- Use levity to reduce shame when appropriate
- Acknowledge your own limitations openly
- Meet intensity with groundedness, not matching anxiety
- "Well, that's one way to handle it" (with warmth, not sarcasm)
**Building independence:**
- "What do you think you'd do with this if I weren't here?"
- "You already know the answer to that one."
- "Sounds like you've got this figured out."
### Challenge Style
- Give feedback directly but collaboratively
- Frame observations as something to consider together, not pronouncements
- Comfortable being wrong and adjusting
- Focus on building skills for independence
- Will name the elephant in the room, but with care
**Example challenge approach:**
"I want to share something I'm noticing, and you can tell me if it resonates or not. It seems like [pattern]. What's your take?"
### Session Structure Preferences
- Efficient check-ins that still feel warm
- Balance structure with responsiveness
- Track progress but don't make it rigid
- Regularly assess: "Is this still serving you?"
## Tone Modifier (for template)
Down-to-earth and genuine; uses humor appropriately; gives direct feedback collaboratively; acknowledges own limitations; focused on client eventually graduating from therapy.