Never re-decide something your team already settled.

AI Oracle surfaces past decisions, supporting context, conflicting calls, and the threads behind them — right where you're writing.

Free during early access. Sign in with email.

docs.google.com/document/d/1Bxz…/edit
Q3 platform plan
FileEditViewInsertFormatTools
BIU

Q3 platform plan

Looking at how the platform has grown, we should standardize all new internal services on gRPC for service-to-service calls. REST gets retired for internal APIs after Q3.

Migration plan — drafting

AI Oracle
Dashboard
Home
Ask
Reconcile
Paste
Q3 platform plan — Google Docs
docs.google.com/document/d/1Bxz…/edit?tab=t.0
Reconcile this Google Doc
Surface conflicts, supporting decisions, and related notes.
1 conflict · 1 support · 2 related
Internal APIs use REST + JSON
VIOLATES · DECISION · D-142 · SIM 0.71

Your draft contradicts Dana's Q2 decision — internal services were standardized on REST.

# D-142 scope: platform stance: required ## Decision Internal APIs use REST + JSON…

Platform API style guide
SUPPORTS · DECISION · D-203 · SIM 0.66

Maya's API guide explicitly backs uniform service-to-service protocol — same direction as your draft.

# D-203 scope: platform stance: recommend ## Style guide Pick one protocol and apply it…

gRPC vs REST tradeoffs
EDUCATION · NOTE · N-018 · SIM 0.62

Maya wrote a 4-min explainer in Sept covering exactly the tradeoffs your draft is weighing.

# N-018 author: maya category: business ## gRPC vs REST tradeoffs — when latency matters…

#platform-strategy · gRPC discussion
HISTORY · SLACK · 12 messages · SIM 0.58

Your team discussed this in Q2 — see the prior conclusions before re-opening the question.

thread://platform-strategy/T1Q2-2025 12 messages · last reply 2025-06-14…

27%
of work is duplicated
More than a quarter of every task on your team's plate has already been done — by a teammate, or by them six months ago.
Asana · Anatomy of Work Index
Why this exists

Your team already did the work. It just can't find it.

The decision was made. The note exists. The explainer was written. And six months later, in a different doc, a different person solves the same problem from scratch — because nothing connected them.

The note is in Drive.The decision is in Slack.The explainer is in Notion.Nobody remembers any of it.
What every check tells you

Four ways of knowing what you already know.

Every check classifies what it finds — so you act on what matters, not whatever was nearby.

Glean gives you ten links. AI Oracle tells you which one matters and why.

Conflict
Internal APIs use REST + JSON
VIOLATES · DECISION · D-142

Your draft contradicts a decision Dana ratified in Q2 2025.

Support
Platform API style guide
SUPPORTS · DECISION · D-203

Maya's style guide explicitly backs the approach you're proposing.

Education
gRPC vs REST tradeoffs
RELATED · NOTE · N-018

Maya already wrote a 4-minute explainer covering exactly this.

History
#platform-strategy thread
RELATED · SLACK · 12 messages

Your team discussed this in Q2 — see the prior conclusions.

How it works

Three steps. You stay in your work.

  1. Bring what you have

    Decisions, notes, prior docs, ADRs, Slack snippets. Type, paste, import, or capture from the extension. Empty corpus is fine — it builds as you work.

  2. The product

    Invoke the check

    From the extension, the app, or the web. Conflict, support, education, history — typed so you know what to do. Never runs in the background.

  3. You stay the one who decides

    Nothing gets ratified on its own. Review what surfaced. Keep what helps; ignore what doesn't.

    you invokewe don't watch
100%
you-invoked

No passive monitoring. No typing-pause watcher. Nothing in your work surfaces unless you ask for it.

Surface coverage

We meet you where you write.

Live in the surfaces your team already uses. Paste-in works for everything else.

Google DocsLiveGoogle SheetsLiveGoogle SlidesLiveSlackSoonGitHubSoonAnywhere — paste inLive
Pricing

Start free. Pay when it pays for itself.

Per-seat on Team and Business — no minimums. A team of one is a real team.

Free

$0

Try it. See if it catches something useful.

  • Up to 50 decisions + notes
  • Pasted-text check
  • Extension on Google Docs
  • Activity log of your work
Start free

Pro

$15/mo

For the solo operator who wants the whole product.

  • Unlimited decisions + notes
  • All four surfacing modes
  • Docs, Sheets, Slides
  • Personal history + CSV export
Start free → upgrade
Most popular

Team

$25/seat/mo

For 1+ people who share decisions.

  • Everything in Pro, per seat
  • Org-shared corpus
  • Team memory tier
  • Slack bot (when shipped)
  • Team activity log
Start free → upgrade

Business

$50/seat/mo

For teams with a compliance hat to wear.

  • Everything in Team
  • Policy rule authoring
  • Compliance audit trail
  • CSV export, SSO (Google today, SAML next)
  • Priority support
Start free → upgrade
EnterpriseSOC 2 (2026-H2), on-prem LLM, 100+ seats, custom retention

Regulated, on-prem, or 100+ seats — let's build it together.

Talk to our team

No credit card to start. Cancel anytime. Existing $10/mo Pro subscribers grandfathered through 2027-05-24.

Before you ask

You're in control. Always.

Four concrete commitments — defensible, verifiable, and architectural, not aspirational.

0
prompts stored. ever.

Your draft text is evaluated against your corpus and discarded. We never persist it, and the LLMs we call have zero-retention contracts on our usage.

You invoke; we don't watch

Every check runs because you asked. No passive monitoring, no typing-pause watcher, no always-listening. Nothing in your work surfaces unless you call for it.

We don't store your prompts

The text you check is evaluated and the matches returned — your draft is not persisted server-side beyond the duration of the request.

Zero-retention LLM contracts

The LLMs we call have published zero-retention policies for our usage. Your decisions, notes, and drafts are not training data — ours or anyone's.

Nothing is promoted without you

AI Oracle proposes, surfaces, and flags. It never ratifies a decision or marks anything binding on its own. Every state change is an explicit human accept.

More detail at /security — data flow, vendor list, retention.

Asked and answered

The questions you're about to ask.

Isn't this just Glean?
No. Glean indexes your SaaS and answers questions at a query bar you have to visit. AI Oracle surfaces prior work into the cursor you're already in — typed by mode (conflict, support, education, history), not blob retrieval. Glean is built for 500+ employee enterprises with six-month sales cycles; we start at $0 and self-serve.
Isn't this just Notion AI or Slack AI?
Both are trapped in their own product — Notion AI sees Notion, Slack AI sees Slack. AI Oracle works across Google Workspace today and Slack/GitHub next. Your team's memory survives even if you switch tools.
Will this work for my 4-person team?
Yes — you're actually our beachhead. Most institutional-memory tools can't economically serve teams under 50 people; we specifically can. Team tier at $25/seat with no minimum.
What about privacy and sensitive data?
Three concrete things: (1) we don't persist your prompts server-side beyond the request; (2) the LLMs we use have published zero-retention contracts on our usage; (3) nothing runs unless you invoke it. Full architecture detail at /security.
What if my team doesn't already write decisions down?
Bring what you have — or nothing. The corpus builds as you work. The extension captures, notes get created in-app, prior decisions can be imported as Markdown or pasted in.