TextWatermark Logo TextWatermark.
Login Sign Up

Encode Hidden Intent Within Any Text.

A mathematically-precise API & MCP server that embeds and decodes watermark payloads using invisible Unicode characters—without changing how your text looks or feels.

Get Your API Key Read The Docs
Text watermarking concept visual

Why TextWatermark?

Steganography

Unicode Precision

Attach invisible Unicode variation selectors to visible characters; the text looks and reads the same.

Reversibility

Lossless Decode

Read the exact payload back after copy-paste between apps, chat and email. Robust mode also reports when the text was cut short or changed.

Integration

Built for Agents

REST API with an OpenAPI schema, a remote MCP server and an Agent Skill. Any language that can send HTTPS works.

Reliability

Transparent Status

Live service health on our public status page. DPAs available on request.

Developer-Centric Toolkit

API Reference

View Full API Reference

Interactive API Console

Configure your request:

POST

Response:

// API response will appear here

Impressed by the demo? Sign up now to integrate TextWatermark with your own content and secure your data.

Create Your Free Account

Straightforward Pricing Plans

Choose the plan that fits your needs. Start for free, scale as you grow.

Free

$0/mo

100 credits to get started

  • 100 one-time credits
  • Fast & robust watermarking
  • 1 API key
Start Free
Most Popular
Starter

$7 /mo

Billed annually ($86/year)

  • 5,000 credits / month
  • Fast & robust watermarking
  • 3 API keys & signing channels
  • Email support
Choose Starter
Pro

$23 /mo

Billed annually ($278/year)

  • 25,000 credits / month
  • Fast & robust watermarking
  • 10 API keys & 20 signing channels
  • Priority support
Choose Pro

Frequently Asked Questions

How does text watermarking work?
We attach invisible Unicode variation selectors to the visible characters of your text. The words, spacing and layout stay exactly the same; the selectors carry your payload (for example a document or recipient id), and the API reads it back later. Normalizing the text, retyping it, OCR or having a model rewrite it removes the selectors, so treat the mark as evidence when it is present, not as proof of origin when it is absent.
Is the watermarking process secure?
It depends on the mode. A plain watermark is hidden, not encrypted: anyone with a TextWatermarking API token can read it, so embed an opaque id (for example doc-8841) and keep the details in your own system. For private payloads, create a signing channel and bind your API token to it; the payload is then encrypted with AES-256-GCM and only tokens on that channel can read it, and changed text fails the check. All API traffic uses HTTPS.
What are the limits on Unicode character embedding?
Any script works as the carrier text, including Turkish, Arabic, CJK and emoji, and the payload can be any UTF-8 string. The limit is length: each mode needs a minimum number of visible characters for a given payload (plain mode needs very little; keyed and robust modes need about 8 visible characters per payload byte). The MCP tool watermark_capacity checks this for free before you spend a credit.
Do you offer uptime guarantees or SLAs?
We don't offer a contractual uptime SLA today. You can see live service health on our status page. If you need specific support commitments for your use case, contact us and we'll tell you honestly what we can provide.