Skip to main content
These examples show how 3ngram works in practice across different roles. You don’t need to type tool names directly. Your AI translates plain language into the right calls. The tool details are shown so you know exactly what happens under the hood.
Project names like acme-saas and client-meridian match the kind of labels you’d use in practice. Replace them with your own.

​
Real-world scenarios

Before diving into tool mechanics, here is how 3ngram fits into different workflows.
Sarah manages five clients, each with its own timeline, deliverables, and decision history. She uses Claude for strategy, ChatGPT for research, and Google Docs for proposals.Without 3ngram: She opens Claude on Monday and spends 10 minutes re-explaining Client A’s situation. She promised Client B a revised proposal by Wednesday but forgot. It was discussed in a ChatGPT session last week. When Client C asks “didn’t we decide to go with approach X?”, she cannot find where that decision was made.With 3ngram: She opens any AI tool and gets a briefing: Client B’s proposal is overdue, Client A has a meeting Thursday, Client C’s decision from two weeks ago is captured with full rationale. She asks “what did we decide about Client C’s migration approach?” and gets an answer with citations from three different sessions.
James runs a B2B SaaS product. Decisions happen in Jira, Slack, Claude, and meetings. His team records calls, but nobody reviews the full transcripts.Without 3ngram: Sprint planning starts and nobody can remember why they scoped the feature this way. A customer commitment from a sales call never made it into the backlog. Three people have different recollections of the same architectural decision.With 3ngram: Decision rationale is captured from the Claude session where it was discussed. Customer commitments are tracked with deadlines. When sprint planning starts, James asks “what are our open commitments for this quarter?” and gets a structured list with sources.
Marcus runs a 12-person startup. He is in every conversation: investor updates, customer calls, product decisions, hiring. He uses AI tools 20+ times per day.Without 3ngram: He makes a strategic decision in a Claude session at 7am, forgets to tell the team, and discovers three weeks later that engineering went a different direction. His investor update is late because he cannot reconstruct what happened this quarter.With 3ngram: His decisions are captured with rationale and tagged to projects. A morning briefing surfaces what is overdue across all workstreams. When it is time for the investor update, he asks “summarize our progress on the three strategic priorities this quarter” and gets a sourced answer from his own decision history.
Priya works across Claude Code, Cursor, and ChatGPT. She makes 30+ architectural decisions per week.Without 3ngram: She starts a Claude Code session and re-explains the auth approach she settled on yesterday. A decision in Cursor about the schema contradicts what she told ChatGPT. Her CLAUDE.md is 200 lines of stale context.With 3ngram: Every AI session starts with her full decision history. GitHub repos are indexed, so the AI can reference actual code alongside her decisions. Recurring commitments auto-advance when resolved.

​
1. Remember + Recall flow

Save a decision, then find it later using semantic search.

​
Save a decision

“3ngram: Remember that we decided to use feature flags for the billing rollout. Tag it to the acme-saas project.”
Your AI calls: 
{
  "tool": "remember",
  "arguments": {
    "topic": "billing-rollout-strategy",
    "content": "Decided to use feature flags for the billing rollout. Lets us control exposure per-customer without a full deploy cycle. LaunchDarkly for the flag provider.",
    "memory_type": "decision",
    "project": "acme-saas"
  }
}

​
Recall it later

In a different session, or a different AI client entirely:
“3ngram: What have we decided about billing for acme-saas?” 
{
  "tool": "recall",
  "arguments": {
    "topic": "billing",
    "project": "acme-saas",
    "brief": true
  }
}
Brief mode gives you one line per memory. Drill into the one you need:
“3ngram: Show me the full details of memory 42” 
{
  "tool": "get_memories",
  "arguments": {
    "ids": [42]
  }
}
Workflow: scan with brief recall, then read what matters. Keeps conversations focused.

​
2. Commitment tracking + resolve

Create a commitment with a due date, check status, and mark it done.

​
Create a commitment

“3ngram: I committed to shipping the onboarding flow for acme-saas by Thursday.” 
{
  "tool": "remember",
  "arguments": {
    "topic": "ship-onboarding-flow",
    "content": "Ship the new user onboarding flow with email verification and workspace setup wizard.",
    "memory_type": "commitment",
    "due_date": "2026-03-20",
    "project": "acme-saas"
  }
}

​
Check open commitments

“3ngram: What are my open commitments for acme-saas?” 
{
  "tool": "commitments",
  "arguments": {
    "project": "acme-saas"
  }
}

​
Check overdue items

After the deadline passes:
“3ngram: Anything overdue?” 
{
  "tool": "overdue",
  "arguments": {
    "project": "acme-saas"
  }
}

​
Resolve it

“3ngram: Resolve the onboarding commitment. Shipped in PR #87.” 
{
  "tool": "resolve",
  "arguments": {
    "memory_id": 67,
    "resolution": "Shipped in PR #87"
  }
}
The commitment moves to resolved status and won’t appear in future overdue checks. The resolution note is preserved for audit.

​
3. Session briefing + debrief

Use built-in prompt templates to start and end sessions with structured context.

​
Start a session with a briefing

“3ngram: Run the briefing prompt”
The briefing prompt orchestrates 7 tool calls behind the scenes: overdue, blockers, commitments, stale_commitments, recall (recent decisions), suggested_context, and status. Your AI composes the results into a structured overview: 
Session Briefing — 2026-03-18

OVERDUE (1):
  [67] Ship onboarding flow — 1 day overdue [acme-saas]

BLOCKERS (1):
  [71] Waiting on design review for settings page [acme-saas]

OPEN COMMITMENTS (3):
  [67] Ship onboarding flow — due: Mar 20 [acme-saas]
  [54] Write webhook API docs — due: Mar 21 [acme-saas]
  [73] Invoice template for freelance-consulting — due: Mar 25

STALE (1):
  [48] Migrate auth to OAuth 2.1 — no activity for 12 days [acme-saas]

RECENT DECISIONS (2):
  [42] Use feature flags for billing rollout [acme-saas]
  [38] Switch from per-seat to usage-based pricing [acme-saas]
This gives you a complete picture of where things stand before you start working. No manual recall needed.

​
End a session with a debrief

When you’re done working:
“3ngram: Run the debrief prompt”
The debrief prompt scans your conversation and extracts:
  • Decisions made — saved as decision memories
  • Commitments created — saved as commitment memories with due dates
  • Blockers identified — saved as blocker memories
  • Items resolved — marked as resolved with notes
Session Debrief — 2026-03-18

EXTRACTED:
  + Decision: Use LaunchDarkly for feature flag provider [acme-saas]
  + Commitment: Update staging env config by Monday [acme-saas]
  + Resolved: Ship onboarding flow — PR #87 [acme-saas]

PERSISTED: 2 new memories saved, 1 commitment resolved.
The debrief captures context you might forget to save manually. Run it at the end of every session, or set up Claude Code hooks to trigger it automatically.

​
Putting it together

A typical daily workflow:
  1. Start with a briefing to see overdue items, blockers, and open commitments
  2. Work normally — save decisions and context as you go
  3. Check commitments mid-day if you need a reminder
  4. Resolve items as you ship them
  5. End with a debrief to capture anything you missed
All of this happens inside your AI chat. No context switching, no separate app required (though the dashboard is there when you want a visual overview).