Codex — Client Document Intelligence

Your protege reads the document.
With your client. On the call.

Codex is the Apex Replicant premium add-on that lets clients upload their own materials (contracts, P&Ls, resumes, pitch decks, plans) and have your protege read, analyze, and discuss them during a live voice session. Per-client. Per-session. Hard-isolated. The expert's KB is permanent; Codex is the conversation-specific intelligence your protege needs for this client, on this call.

Add Codex to your account→Talk to sales about vertical deployments→
Your KB is one thing. Your client's document is another.

Why Codex exists.

Your knowledge base is your protege's long-term memory: methodologies, frameworks, books, transcripts, the way you think about the work. It's permanent. It shapes every conversation.

But the specific contract your client is staring at right now? The P&L they're trying to understand? The resume your protege is helping them edit? That's not part of your protege's memory and it shouldn't be. It belongs to that client, on that session, and to no one else.

Codex is the surface that handles those documents.

Two surfaces. One protege.

How Codex is different from your Knowledge Base.

Knowledge Base
Codex
  • Who uploads: the expert
    Who uploads: the client
  • When: once, refined over time
    When: per session, with the client's consent
  • Scope: all conversations, all clients
    Scope: this conversation, this client, this session
  • Retention: permanent
    Retention: 90 days inactivity, soft-delete with email warning
  • Use case: teach my protege how I think
    Use case: let my protege read my client's contract on the call
  • Pricing: included in every plan
    Pricing: $49/mo or $399/yr add-on per expert
Three steps. Inside the live session.

How it works.

  1. Client uploads

    Before or during the session, the client drops in their document: PDF, DOCX, TXT, MD, HTML, audio (MP3, M4A, WAV, OGG, WebM), video (MP4, M4V). No expert curation required.

  2. Protege reads and analyzes

    During the live voice call, the protege reads the document in real time, references it naturally in the conversation, and answers questions grounded in its specifics.

  3. Session ends, Codex closes

    The document stays available to that client for that protege's future sessions; the 90-day retention clock starts on inactivity. No cross-client leakage. Ever.

Per-expert, per-protege, per-client, per-session

4D Knowledge Scoping.

Every Codex retrieval is scoped by four dimensions: which expert, which protege, which client, which session. One client's documents cannot leak into another client's conversation. The isolation is enforced at retrieval time, in the database, with row-level security: not as a prompt-level "please don't share."

This is the same architecture pattern Sentinel uses for B2B knowledge isolation. It's why non-technical buyers can hand Codex documents to a protege without an enterprise IT review. See /security for the full architecture write-up.

Live document analysis, in the expert's voice

What you can do with Codex.

  1. Attorneys

    Your client uploads a contract before the session. Your protege reads it, flags the clauses you'd flag, and answers their questions in your voice for the full session, without you on the call.

  2. Financial advisors

    Your client drops in a P&L. Your protege walks them through the line items, contextualizes them against the framework you teach, and routes the high-value follow-ups to your calendar.

  3. Executive coaches

    Your client uploads a 360 review. Your protege works through the implications with them in the way you would, using your methodology, with no risk that another client ever sees what this client shared.

Codex for clients

Your private space with your expert.

If your expert has Codex enabled, you have a secure place to share documents with them: contracts, P&Ls, resumes, plans, anything you'd normally walk into a meeting with. Drop the file in. Their protege reads it. Your session is anchored in your actual situation, not a generic conversation.

Your Codex space is yours. It's scoped per-client, per-session: per-expert per-protege per-client per-session. No other client of that expert ever sees what you uploaded. The expert can't read or export your documents directly. The documents only surface through the protege, during your session.

For a plain-English walkthrough of how a session works and what to expect, see /for-clients and /sessions.

Premium add-on. Two ways to pay.

Pricing.

Both tiers include unlimited client document uploads, unlimited Codex-aware sessions, 90-day retention, and the 4D scoping architecture. No per-document fee. No per-session fee.

  • Monthly

    Monthly

    $49/month

    Per expert. Cancel any time.

    Best for: Experts validating Codex with an existing client roster.

    • Unlimited client document uploads
    • Unlimited Codex-aware sessions
    • 90-day inactive-document retention
    • 4D Knowledge Scoping (RLS-enforced isolation)
    • Cancel any time
    Add Codex monthly
  • Most Popular

    Annual

    $399/year

    Per expert. Two months free vs monthly.

    Best for: Experts whose practice depends on document-driven sessions.

    • Everything in Monthly
    • Two months free (~32% savings)
    • Annual billing, single invoice
    Add Codex annual
FAQ

Common questions.

What file types does Codex accept?
PDF, DOCX, TXT, MD, HTML, audio (MP3, M4A, WAV, OGG, WebM), and video (MP4, M4V). Same set the platform supports for KB uploads.
What happens to my client's document when the session ends?
It stays available to that specific client, in that specific protege, for future sessions. After 90 days of inactivity we soft-delete it with an email warning to the client. Your KB is untouched.
Can my protege use Codex documents alongside my KB?
Yes. During a session the protege pulls from both: your permanent KB (your methodology, your voice, your frameworks) and the client's current Codex document (the specific contract, P&L, or resume in front of you). The two are layered together at retrieval time but stay isolated at storage.
Does PII get scrubbed?
All ingested content runs through our PII detection and removal pipeline before storage. The original document is preserved at the file layer for review; the indexed embeddings are PII-stripped.
Can clients export or delete their Codex documents?
Yes. Clients control their own documents through the client-side surface. Experts cannot read or export a client's Codex documents directly; the documents only surface through the protege during a session.
Document intelligence in your voice

Let your protege read
with your client.

$49/month. $399/year. Unlimited client documents. The only voice AI on the market that does this.

Add Codex to your account→Read the architecture deep-dive→