KINDI masks PII in text before you send it to an LLM, and lets your client decrypt the response on the way back.

Paste your mk_live_… key once below. It is stored in this browser only (never sent as a cookie) and unlocks the live Run panes on the mask-and-decrypt and redact example pages.

The server never persists token mappings. For /mask, decrypting the masked text is your responsibility; see the mask-and-decrypt example for runnable Python and TypeScript snippets. The LLM Proxy is server-side: there KINDI masks and unmasks for you and the mapping lives only in the request (then discarded), still never persisted.

Create an API key on the API keys page.
POST /api/v1/mask with your text and the bearer.
Decrypt the envelope client-side.

Try it in the playground →

Using an AI coding agent?

If you're driving the integration with Claude Code, Cursor, GitHub Copilot, Codex CLI, Windsurf, Aider, or any other agent that reads a repo-level instructions file, install the KINDI skill once and your agent learns the API surface, the exact crypto parameters, the token formats, and the anti-patterns to avoid:

12345# Claude Code (project-scoped)
curl -fsSL https://docs.kindi.me/kindi.agents.md >> CLAUDE.md

# Codex CLI / Zed / cross-tool default
curl -fsSL https://docs.kindi.me/kindi.agents.md >> AGENTS.md

See the install command for your tool →

Three machine-readable surfaces for agents

Surface	URL	When to use
Skill file, pinned guidance	`/kindi.agents.md`	Paste-once into your repo so the agent always has KINDI context
Docs index, links to every page	`/llms.txt`	Lean; the agent fetches individual `.md` pages on demand
Bulk download, every page as a `.md` in one archive	`/docs.zip`	Pre-pull for offline grep; or pre-load when the agent has lots of context budget

Every page on this site is also available as raw markdown: append .md to the URL (e.g. docs.kindi.me/reference/mask.md).