JSNexis for JavaScript and AI apps
Find names, emails, card numbers, health and ID numbers in your text, then redact, mask or encrypt them before the text reaches an LLM, a log or a third party. Detection follows the regulation you choose. Use it from your code, your terminal or your AI assistant.
Three packages, one engine
Pick the one that fits how you work. They all use the same Nexis API key and the same detection rules.
SDK and CLI
@intelation/nexisClean text in your own Node code, or straight from the terminal. Zero dependencies.
npx @intelation/nexis redact "Email john.smith@example.com"
# -> Email [REDACTED]MCP server
@intelation/nexis-mcpLet Claude, Cursor or VS Code find and remove personal data when you ask.
claude mcp add nexis \
--env NEXIS_API_KEY=your-key \
-- npx -y @intelation/nexis-mcpAI SDK middleware
@intelation/nexis-aiOne line removes personal data from every prompt in a Vercel AI SDK app.
const model = withNexis(
openai("gpt-4o"),
new Nexis()
);Protect the prompt going in, restore the answer coming out
Wrap your model once and every prompt is cleaned before it leaves your server. Use encryption and the model works on placeholders, while your user still sees the real names and numbers in the answer.
withNexisReplaces each value with its type.
Pick a regulation
Choose a country and regulation, such as UK GDPR or HIPAA. Nexis looks for the personal data that regulation covers, including local IDs like NHS, National Insurance and Social Security numbers.
Choose what happens to it
Seven methods: redact, replace (with the type), mask, hash, tokenization, pseudonymization (realistic fake values), or encryption so you can restore it later. Or just detect it and leave the text as it is.
Fine-tune when you need to
Process every type of personal data, only the types you pick, or everything except some. Add an allow list of values to keep and a deny list of values to always hide, like a project code name.
Built for the people who build, and the people who sign off
Developers and AI teams
- Free API key, one install, one line of code
- Works with any AI SDK provider: OpenAI, Anthropic, Google and more
- A CLI for scripts, pipelines and quick checks
- An MCP server so your coding assistant can clean data on request
- Plain-English errors that say what to do next
DPOs, privacy and compliance leads
- Detection rules tied to a specific regulation and country
- Fails closed: if Nexis can't be reached, the AI call is stopped, not sent unprotected
- Submitted text is not stored by default
- Encryption keys stay with you; the MCP server never passes them through the model
- Audit logs of anonymization activity in the Nexis dashboard
Regulations and jurisdictions supported
Your plan sets how many regulations you can use.
UK GDPR
United Kingdom
GDPR
Germany
English and German textHIPAA
United States
PIPEDA
Canada
Privacy Act
Australia
Privacy Act
New Zealand
How to get started
Get your free API key
Sign up for Nexis with your work email. Approval usually takes minutes. Then create a key in the dashboard under API Integration.
Install a package
Pick the SDK and CLI, the MCP server or the AI SDK middleware from npm. Node 18 or later.
Set your key and go
Set
NEXIS_API_KEYin your environment. Every call uses the first regulation in your plan unless you choose another.
npm install @intelation/nexis-ai aiimport { generateText } from "ai";
import { openai } from "@ai-sdk/openai";
import { Nexis, withNexis } from "@intelation/nexis-ai";
const model = withNexis(openai("gpt-4o"), new Nexis());
await generateText({ model, prompt: "Summarise the ticket from john.smith@example.com" });
// The model receives: "Summarise the ticket from [REDACTED]"Works with any provider and with generateText, streamText and generateObject. AI SDK v5, v6 and v7.
Encryption and decrypt
Encryption is the only method you can reverse. It needs a Fernet key in NEXIS_ENCRYPTION_KEY. Generate one with either command, and keep it safe:
node -e "console.log(require('crypto').randomBytes(32).toString('base64url') + '=')"python -c "from cryptography.fernet import Fernet; print(Fernet.generate_key().decode())"Restore the values with the same key: nexis.decrypt() in the SDK, decrypt in the CLI, the decrypt_text MCP tool, or restore: true in the AI SDK middleware. Ask your AI model to keep the <<ENC:...:ENC>> markers unchanged in its answer.
Questions
Can't find your answer? Talk to us.
Do I need a separate API key for each package?
Which package should I use?
nexis-ai. Any other Node code or a script: nexis. Want Claude, Cursor or VS Code to clean data on request: nexis-mcp. Working in n8n: see Nexis with n8n.Does Nexis store my text?
What happens if Nexis can't be reached?
Does it add cost to every AI call?
Will it catch every piece of personal data?
detect to check results, add values you always want hidden to the deny list, and review high-risk flows.Can I get the original data back?
Are images and files scanned?
Can I use Python or another language?
https://api.nexis.intelation.com.