How it works

Log decisions, store them in your project folder, load them automatically in AI chats.

Step 1

Log decisions

Free: Hit Cmd+Shift+L, pick a template, fill in the blanks.Pro: Continuity detects changes (git commits, file saves) and auto-generates decisions for you to review.

  • Free: 5 templates, manual logging
  • Pro: Auto-detects changes via git hooks, file watchers, AI session tracking
  • Tags extracted automatically, duplicates filtered out

# Log decisions

## Free: Manual logging
Hit Cmd+Shift+L, pick a template:
• "Why choose X over Y?"
• "Why implement X this way?"
• "Why refactor X?"

Example:
[You press Cmd+Shift+L]
Question: "Why use Zod for validation?"
Answer: "Zod for TypeScript-first design..."
✅ Logged with auto-extracted tags

## Pro: Auto-detection
$ git commit -m "Add Zod for validation"

[Continuity detects change]
→ Generates Q&A automatically
→ You review and approve
→ Gets logged

Detects via: git hooks, file saves, AI sessions

# Storage in your project

.continuity/
├── decisions.json          # Your decisions
├── SESSION_HANDOFF.md      # AI context file
├── delta-snapshot.json     # Change tracking
└── embeddings.json         # Semantic search (Pro)

# Example decision:
{
  "id": "1761121236686-bgocje",
  "question": "Why use Zod over Joi?",
  "answer": "Zod for TypeScript-first design...",
  "tags": ["validation", "typescript"],
  "timestamp": "2025-10-22T08:20:36.686Z",
  "files": ["package.json"]
}

✅ Plain JSON—readable, editable
✅ Commit to git
✅ Stays local—never hits the cloud
✅ SESSION_HANDOFF.md updates automatically
Step 2

Store in your project

Everything goes in a .continuity/ folder. Plain JSON files you can read, edit, and commit to git. SESSION_HANDOFF.md updates automatically with your latest project context.

  • Readable JSON—no database, no lock-in
  • SESSION_HANDOFF.md generates: architecture summary, recent commits, decisions, git status
  • Commit to git, share with your team
  • Pro: Semantic search finds relevant decisions faster
Step 3

Load automatically in AI

Open Claude, Cursor, or any AI tool. SESSION_HANDOFF.md loads automatically. Your AI already knows your architecture, recent commits, and last 26 decisions. No copy-paste.

  • Loads project context in under 500ms
  • AI can log new decisions, search old ones, track what you've documented
  • Works with Claude (Desktop/CLI/Code), Cursor, Cline, Roo Code, Copilot
  • Uses MCP protocol—standard across AI tools

# Auto-loads in AI

[Open Claude, Cursor, any AI tool]

SESSION_HANDOFF.md loads automatically:
✅ Project architecture & tech stack
✅ Recent commits & file changes
✅ Last 26 decisions
✅ Git status & current branch

You: "How should we validate API inputs?"

AI: "You're using Zod for validation (logged 3 days
     ago). Here's a schema that matches your style..."

# AI can call these tools:
─────────────────────────────────────
• log_decision(question, answer, tags)
• search_decisions(query)
• get_compliance_metrics()
─────────────────────────────────────

✅ No context loss between sessions
✅ Works with Claude, Cursor, Cline, Copilot
✅ Loads in <500ms

# Verify it's working

Ask your AI:
"Do you have access to Continuity?
What memory tools are available to you?"

Expected response:
✅ Yes, I have Continuity installed

Available tools:
• log_decision(question, answer, tags)
• search_decisions(query)
• get_compliance_metrics()
• SESSION_HANDOFF.md loaded (14KB)

I can see:
- Your project structure
- Recent commits
- Tech stack
- 47 past decisions

If AI says "I don't have those tools":
→ Check VS Code Output → Continuity
→ Reload VS Code (Cmd+Shift+P → Reload Window)
→ Open a project folder (Continuity activates per-folder)
Step 4

Verify installation

Ask your AI if it has Continuity. It'll confirm the tools it has access to and show you what context it loaded.

  • 30-second verification check
  • AI lists available tools and loaded context
  • Instant feedback if something's wrong
  • Builds confidence that it's actually working