/docs/Documentation hub
Canonical overview for builders, security reviewers, and buyers.
Documentation for verifiable agents
Build agents that can prove what they remember.
CORTEX docs are organized around the production path: install locally, admit facts through guards, verify the ledger, expose a tenant-scoped API, and wire MOSKVbot-style agent memory without trusting raw chat transcripts.
Quickstart
Install, initialize the ledger, store a decision, query it, and verify continuity.
02API guide
Memory writes, hybrid search, tenant scope, checkout handoff, and operator endpoints.
03Security
Threat boundaries, tenant isolation, secret redaction, hash continuity, and audit posture.
04MOSKVbot
Reference architecture for a bot with a portable, auditable memory passport.
Operator map
The docs follow the same path as the write pipeline.
CORTEX is easier to understand when the documentation mirrors the system contract: proposal, validation, ledger, persistence, retrieval, export.
Propose
An agent proposes a fact, decision, observation, or handoff state.
Guard
Schema, tenant scope, taint/provenance, redaction, and deterministic validation decide admission.
Persist
Accepted facts are written transactionally with hash-chain continuity and explicit failure semantics.
Verify
Operators can query, replay, export, and inspect evidence when an incident or audit happens.
Core routes
Public surfaces, mapped to actual docs.
These paths are now stable entry points for the landing footer, external links, and future search indexing.
/quickstart/First proof
The shortest path from install to verified memory.
/reference/API reference
Routes, payloads, tenant scoping, and integration examples.
/security/Security posture
Isolation, audit continuity, encryption notes, and operational guardrails.
/architecture/MOSKVbot architecture
How CORTEX becomes the verifiable memory passport for a flagship agent.
/changelog/Evolution
Release notes and roadmap evidence for product maturity.
Deep documentation map
What builders, auditors, and operators need next.
The docs are now organized as a product surface, not a dump of links: each path answers who uses it, what evidence it creates, and where the operational boundary sits.
/crypto-tax/Crypto + NFT audit evidence
Wallet graph, transaction classification, cost-basis support, custody gaps, NFT royalties, and advisor-ready exports.
/blog/Research and positioning
Essays on verifiable memory, agent trust, MOSKVbot, fiscal evidence, and why CORTEX is not just recall.
/architecture/MOSKVbot reference architecture
How to build an assistant with a portable memory passport and proof anchors.
Good-use checklist
Use CORTEX when memory can become liability.
CORTEX is strongest when a future reviewer may ask why the system believed something, who had access, and whether evidence was altered.
Good use
Agent decisions, crypto/NFT audit packs, incident reconstruction, compliance evidence, long-running research, and cross-agent handoffs.
Weak use
Disposable chat context, simple cache lookups, temporary UI state, or low-risk drafts that never become durable operational state.
Buyer
Technical founder, security lead, AI platform engineer, compliance owner, fiscal advisor working with onchain evidence.
Proof
Hash chain, Merkle checkpoint, source manifest, tenant scope, guard outcome, exportable evidence pack.
$ pip install cortex-persist
$ cortex init
$ cortex store --type decision "Use tenant-scoped memory with ledger verification"
stored fact | tenant=default | hash=sha256:...
$ cortex trust-ledger verify --full
ledger intact | checkpoint verified | evidence readyDocumentation standard
No vague memory claims. Each page should expose the contract, the command, the API shape, the failure mode, and the audit implication. That is how CORTEX separates itself from generic agent memory tools.