Build with Claude. Audit with Codex. Optimize with Grok. One privacy-first brain.

Every model. One memory. You hold the keys. Scoped context across the AIs you already pay for — BYOK, audit log, revoke anytime.

Same pack. Three roles. No re-brief.

Works with Claude, Codex, Grok, and Kimi as user-connected tools. Tokens stay on your bills.

Build

Implementation and writing. The vault briefs the model with the project pack only.

Audit

Rigor and review on the same memories. Switch models without pasting history.

Optimize

Sharpen and list alternatives. Still the same aperture. Revoke the session when done.

Aperture in. Revoke out.

We cannot stop a model from seeing tokens we send. We do prevent raw vault access, standing keys into your life, and silent hoarding of the graph.

Scoped packs

minimal · project · personal. Cap size. Exact pack is audited.

Import is a queue

Extract, review, then commit. No silent vacuum of chat exports.

You hold the keys

AES-256-GCM at rest. HTTP desk is bearer-gated. No cookies.

Plans

Subscription is vault + permissions + connectors. Model tokens stay BYOK.

PlanUSDIncludes
Free$01 project, local vault, mock + 1 model, limited MCP
Pro$15/mo ($144/yr)Unlimited projects, MCP, importers, audit, quota desk, multi-model
Pro+$29/moHeavy daily use, failover rules, priority path
Team$24/user/mo (min 3)Shared vaults, roles, admin audit export