3ngram as Your Onboarding Doc

New team member? New AI session? Full context in one call.

Onboarding docs go stale. Tribal knowledge stays tribal. 3ngram gives any new person, or any new AI session, the full accumulated context: decisions, commitments, patterns, and project history. Current as of the last conversation.

1. Onboarding docs are outdated before they're finished

You write a 'Getting Started' guide. By the time the next hire reads it, two services have been renamed, the deployment process changed, and the architecture section describes last quarter's design.

3ngram: Compile context for any project and get a live snapshot: recent decisions, active commitments, current blockers, indexed documentation. Always reflects the actual state because it's built from living memory, not a static doc.

2. Every AI session starts from zero

You open a new Claude or ChatGPT session. It knows nothing about your project, your decisions, your constraints. You spend the first 10 minutes re-establishing context, or worse, you skip it and get generic advice.

3ngram: Session briefings pull open commitments, recent decisions, blockers, and suggested context automatically. Your AI picks up where the last session left off. No re-explaining, no context loss.

3. Handoffs lose everything

Someone leaves the project. Their context leaves with them. What they decided, why they made those choices, what they were working on: all locked in their personal AI chat history or their head.

3ngram: Decisions and context are captured in shared memory. The handoff prompt generates a structured JSON export of all open commitments, recent decisions, and active blockers. The next person's AI starts with everything the previous person knew.

See it in action

Here's what 3ngram returns when you use it as your onboarding doc.

Project Context

>Onboard me to the Beacon project

Project Context: Beacon

Overview:Beacon is a customer health monitoring dashboard. Stack: Next.js + FastAPI + PostgreSQL. Deployed on Vercel (frontend) and Railway (backend). Two active contributors: you and Marcus. Started Nov 2025, currently in beta with 12 pilot customers.

Recent Decisions:Switched from WebSocket to SSE for real-time health scores (Mar 15) -- simpler to deploy, good enough latency for 30s refresh intervals.Adopted row-level security for multi-tenant isolation (Mar 2) -- every query scoped to tenant. No more WHERE clauses in application code.Dropped the mobile app from v1 scope (Feb 20) -- responsive web covers 90% of use cases. Revisit post-launch.

Active Commitments:Launch public beta -- due Apr 18Write API docs for webhook integrations -- due Apr 12Fix false-positive alerts on low-traffic accounts -- due Apr 10

Blockers:Stripe integration blocked on legal review of data processing addendum. Expected resolution: Apr 8.

Key Contacts:Marcus (backend lead), Julia (design), Raj (sales, enterprise pilots)

FAQ

How does a new team member get started?

They connect 3ngram via MCP and run a project context prompt. They immediately get recent decisions, active commitments, blockers, and indexed documentation for whatever project they're joining.

Can I export context for someone without 3ngram?

Yes. The handoff prompt generates a structured JSON export with all open commitments, recent decisions, and blockers. You can share it as a document or paste it into any AI tool.

Does this work across AI tools?

Yes. 3ngram is an MCP server that works with Claude, ChatGPT, Cursor, Windsurf, and any MCP-compatible client. Same memory, same context, regardless of which tool your team uses.

Context That Transfers, Not Documents That Decay

3ngram gives every new session and every new team member the full accumulated context. Decisions, commitments, and project history, always current.

Request access