Skip to main content

​
Sign up

  1. Go to app.3ngram.ai and create an account
  2. Complete onboarding to set up your workspace
Once signed in, you can use the dashboard to browse memories, commitments, and decisions. The core workflow, though, is inside your AI tool: brief, work, capture decisions, and follow up later.

​
Connect an MCP client

3ngram works through the Model Context Protocol (MCP). Connect any MCP-compatible AI tool to give it persistent memory across sessions. For the current step-by-step setup by client (Claude.ai, Claude Code, Cursor, ChatGPT), use the canonical guide: That page is the source of truth for UI labels, CLI commands, and transport details.

​
Save your first memory

With 3ngram connected, ask your AI assistant to remember something. Prefix with “3ngram:” so the AI routes to 3ngram instead of its own native memory:
“3ngram: Remember that we chose PostgreSQL for the database”
3ngram stores it with the right type, project tag, and scope automatically. Try saving a commitment with a due date:
“3ngram: Remember this commitment due Friday: finish the API review”
Your AI will confirm the save:
✓ Remembered “API review” as commitment [due: 2026-03-14] (id: 42)
Many AI tools (Claude Code, Cursor, ChatGPT) have their own built-in memory. Always prefix with “3ngram:” or “in 3ngram” to make sure your memory is saved to 3ngram, not the tool’s local store. See Memory Tools for more examples.

​
Recall memories

Ask naturally:
“3ngram: What do you remember about the database decision?”
“3ngram: What commitments do I have this week?”
“3ngram: Show me my open blockers”
3ngram uses semantic search, so you don’t need exact keywords. 3ngram returns matching memories with context:
database-decision — decision — 3d ago [project: engram] We chose PostgreSQL for the main database because of pgvector support for semantic search and mature tooling.

​
Run a briefing

Start your next session with a briefing to pick up where you left off:
“3ngram: Run my briefing”
This gathers open commitments, blockers, overdue items, and recent decisions into a structured summary. A briefing pulls together your open items:
Open Commitments (2):
[42] Finish the API review (1 day) [project: engram]
[38] Update deployment docs (5 days) [project: engram] Blockers (1):
[35] Staging SSL certs not provisioned (3 days) [project: engram] No overdue items ✓

​
Use the core workflow

The fastest way to feel the product value is:
  1. Start your session with 3ngram: Run my briefing
  2. Save decisions and commitments while you work
  3. End the session with 3ngram: Run my debrief
That turns 3ngram from passive memory into an actual follow-through loop.

​
Explore your tools

Once connected, ask your AI assistant to show you what 3ngram can do:
“3ngram: What tools do you have available?”
The AI will list all available MCP tools and explain what each one does. This is the fastest way to discover capabilities like search_content, get_schedule_history, or configure_scope that you might not find in docs alone.

​
Custom instructions

For the best experience, add this to your AI client’s custom instructions or system prompt: 
When I mention "3ngram" or ask about memories, commitments, decisions, or blockers,
use the 3ngram MCP tools. Start sessions with the briefing prompt and end with the
debrief prompt. Always route memory operations to 3ngram, not your built-in memory.
Where to paste this depends on your client:
  • Claude Code: Add to your project’s CLAUDE.md file
  • Claude Desktop: Settings > Custom Instructions
  • Cursor: Settings > Rules for AI

​
Next steps

Connect more clients

Set up Claude Code, Cursor, or ChatGPT.

Memory tools

Learn all tools and how to use them naturally.

Prompt templates

Automate briefings, debriefs, and project context.

Dashboard

Review commitments, nudges, and insights in the web UI.