ogikol
Pillar I · Privacy Vault

Sensitive datanever leaves your tenancy.

An API-first vault that detects PII, tokenizes it into safe surrogates, and encrypts the originals under keys you control. Sit it in front of any model, any analytics tool, any partner — they see structure, never secrets.

Capabilities

Four operations. One contract.

Detect, tokenize, encrypt, govern — exposed behind a single, consistent API. Compose them; chain them; run them inline on every request crossing a trust boundary.

Detect

Find PII anywhere it hides.

Six detection layers run in parallel: pattern + checksum, schema + tag, NER, vision (stamps/signatures/faces), context inference, and your own custom rules. Structured rows, free text, scanned images — nothing slips through.

Tokenize

Replace secrets with safe surrogates.

Format-preserving tokens that look real to your downstream tooling. SSN-shaped strings stay SSN-shaped; account numbers stay numerically valid. Tokens are scoped to your tenancy and impossible to reverse without your keys.

Encrypt + split

Originals locked, mirrors emitted.

Originals are encrypted at field level under keys held in your KMS, then stored in one repository. The tokenized mirror — same shape, no secrets — lives in another. Compromise of one store gives an attacker nothing useful.

Govern + audit

Sign every action.

Every authorize, access, tokenize, and detokenize event is recorded and cryptographically signed. Hand a regulator a verifiable export of any time window in seconds. Drag a timeline in your console; rebuild any moment.

Detection

Six layers, run in parallel.

No single detector finds everything. Together they cover structured fields, free text, scanned images, and inferred identifiers — and you can extend the bottom layer with rules specific to your domain.

Patterns + checksums
regex · Luhn · ABA · MRZ · IBAN · custom
Schema + tags
column types · field tags · DBT marts
Named entities
names · orgs · locations · dates · IDs
Vision
stamps · handwriting · signatures · faces
Context inference
neighborhood + co-occurrence
Custom rules
your domain dictionary · YAML rules
API

One verb in. A safe surrogate out.

Tokenize and detokenize are single-call, idempotent, and schema-aware. Wrap them in your existing data layer or call them inline from any service.

POST /v1/tokenize
batch · idempotent
curl -X POST https://api.logikol.com/v1/tokenize \
  -H "Authorization: Bearer $LOGIKOL_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "fields": [
      { "key": "name",    "value": "Maria Lopez",    "kind": "name" },
      { "key": "ssn",     "value": "412-55-9180",    "kind": "ssn"  },
      { "key": "account", "value": "8801-4421-3092", "kind": "account" }
    ],
    "policy": "default"
  }'
200 OK · format-preserving tokens
reversible by policy
{
  "tokens": {
    "name":    "[name_a3f]",
    "ssn":     "[ssn_91c]",
    "account": "[acct_4de]"
  },
  "refs": {
    "name":    "vault://acme/customers/9f1c2.../name",
    "ssn":     "vault://acme/customers/9f1c2.../ssn",
    "account": "vault://acme/customers/9f1c2.../account"
  },
  "audit_id": "evt_2026_05_05_4z9k1q"
}
Primitives

Four building blocks.

Everything Logikol does is composed from a small set of primitives. Learn these four and the rest of the platform follows.

Tokens

Format-preserving surrogates with stable shape (SSN-like, account-like, name-like). Reversible only inside your tenancy with a key fetched from your KMS at request time.

Vault references

Pointers to the encrypted original. Travel everywhere safely; the data behind them never leaves your perimeter without explicit detokenize-by-policy.

Policies

Per-field rules deciding who can read, who can detokenize, when, and from where. Expressed as compact YAML; enforced at every API boundary.

Audit chain

Append-only signed log of every authorization and access. Hash-chained, exportable, queryable. The receipt for every interaction with sensitive data.

Integrations

Plugs into the stack you already run.

KMS providers, storage, identity, observability — Logikol meets your stack where it is. No bespoke key-handling layer, no proprietary identity, no “please rip out your existing monitoring”.

KMS
  • AWS KMS
  • GCP Cloud KMS
  • Azure Key Vault
  • HashiCorp Vault
  • On-prem HSM
Storage
  • S3 / GCS / Azure Blob
  • Postgres
  • Snowflake
  • BigQuery
  • Object stores you provision
Identity
  • Okta
  • Azure AD
  • Google Workspace
  • OIDC
  • SCIM
Observability
  • Datadog
  • Grafana
  • New Relic
  • Splunk
  • OpenTelemetry
Pair it with extraction

Document Intelligence + Vault = parse and protect in one call.

The same API can extract a document and tokenize its sensitive fields in a single sync request. The model that consumes the output sees structure, never secrets.