Documentation

Everything to get started and integrate the control-plane.

GET STARTED Quickstart Activate your first tenant and make your first protected call. REFERENCE API Reference Anonymization, token and tenant management endpoints. CONCEPTS Policies & Modes How detection, anonymization and de-anonymization work. CLOUD Cloud Provisioning Tenant lifecycle: from checkout to active tenant.

Quickstart: first call in 3 steps

From active plan to your first anonymized response.

Generate an API key

In the Cloud dashboard → Customer Keys create a new key. Copy it: it is shown only once.

Make your first call

Send a POST request with the text to protect and the context type.

curl -X POST https://api.pseudora.cloud/v1/anonymize \
  -H "Content-Type: application/json" \
  -H "X-Api-Key: pii_cust_YOUR_KEY" \
  -d '{"text":"Mario Rossi, CF RSSMRA80A01H501U","context_type":"generic"}'

Read the response

The text field contains the protected text. The mappings array links each token to the detected type.

{
  "text": "[PERSON_1], CF [FISCAL_CODE_1]",
  "mappings": [
    {"token": "[PERSON_1]",      "type": "PERSON"},
    {"token": "[FISCAL_CODE_1]", "type": "FISCAL_CODE"}
  ]
}

Connect Pseudora to your MCP client

The client connects to the managed gateway; authentication, tenants, ACLs, quotas and auditing remain enforced by Pseudora Cloud.

VS Code: automatic setup

The extension provides @pseudora protected chat and automatically registers MCP tools for Agent workflows. It uses the workspace account, team, document type, and mode without an API key or mcp.json entry.

Sign in

Select “Pseudora: Sign in” in the status bar or run Pseudora: Sign In from the Command Palette.

Select a team in VS Code

The browser only authenticates the account. After the callback, VS Code opens the team picker for the current workspace.

Configure the workspace

Select Pseudora in the status bar and choose document type, mode, and protected or restored AI responses. You can disable only MCP while keeping editor protection active.

Protect prompts

Start the request with @pseudora: the extension anonymizes before invoking the selected model. For Agent workflows, enable MCP tools separately in Configure Tools.

Verify the server

Run MCP: List Servers from the Command Palette and check that Pseudora is Enabled.

Download the VS Code extension

Other MCP clients: manual setup

Create a team key

Open Cloud → Customer Keys and generate a key for the tenant you want to use.

Add the remote server

Configure the MCP endpoint only. There is no repository to clone or local process to run.

Authorize the client

Send the key as a Bearer token. Pseudora attributes requests and usage to the tool and team.

Remote configuration

{
  "mcpServers": {
    "pseudora": {
      "url": "https://mcp.pseudora.cloud/mcp",
      "headers": {
        "Authorization": "Bearer pii_cust_YOUR_KEY"
      }
    }
  }
}

Keep the key in the client credential store. An MCP tool cannot intercept a prompt already received by the host model.

What the PII engine does

Every capability exposed by the core, governable from the control-plane via API.

Anonymization & De-anonymization

Single-text anonymization

›Context + tag or surrogate mode
›Inline policy for runtime override
›Dry run: analysis without persisting the mapping
›include_entity_values (admin only)

anonymize.py

Batch anonymization

›Configurable maximum limit
›Batch-level defaults + per-item override
›Per-item output with processed/failed status

anonymize.py

De-anonymization

›Restore original text from token/mapping
›Audit and usage tracking

deanonymize.py

Detection & protection modes

Multilayer detection pipeline

›Presidio + spaCy
›Privacy Filter
›AI4Privacy
›Regex from database
›Overlap resolution by layer priority

README.md

Protection modes

›Tag mode with stable tokens
›Surrogate mode: realistic, format-preserving fake data
›Deterministic surrogates for the same context

README.md

Hierarchical policy system

Policy system

›Context types, domain policies, PII type registry
›Precedence: inline request > domain policy > type default action

policy-system.md

Context Types — CRUD + versioning

›List, create, update, delete
›Version history per context type

context_types_router.py

Domain Policies — CRUD + versioning

›Upsert policy per domain: protect / keep / surrogate
›Policy version history

domain_policies_router.py

Governable PII type registry

›default action, faker strategy
›reversible, enabled

pii_types_router.py

Runtime detection tuning

Regex patterns — runtime CRUD

›Manage regex patterns from DB without restart

regex_patterns.py

Reclassification rules — CRUD

›Post-detection reclassification rules

reclassification.py

Denylist — CRUD

›Handle recurring false positives

denylist.py

Presidio context words — CRUD

›Context vocabulary to improve detection

presidio_context.py

Runtime, languages & access

Runtime languages & NER models

›List known languages
›Async spaCy model install + status
›Set default language

languages.py

API key management with roles

›Roles: admin, service, auditor
›Create / list / revoke key
›Tenant-aware scoping for tenant admins

identity.py