Developers

Connect in 30 seconds

One click above, or point any MCP client at the Clean.tools server and every tool below shows up ready to call. Pick your client:

Claude Code

claude mcp add --scope user --transport http clean-tools https://mcp.clean.tools/mcp

Claude Desktop / Claude.ai (Settings, Connectors, Add custom connector)

https://mcp.clean.tools/mcp

Cursor (mcpServers)

{"mcpServers":{"clean-tools":{"url":"https://mcp.clean.tools/mcp"}}}

npm (stdio)

npx @clean-tools/mcp

Any other MCP-compatible client: point it at https://mcp.clean.tools/mcp (streamable HTTP, authless, no key).

VS Code, Windsurf, Cline, Zed, Continue, Gemini CLI, JetBrains, Warp, Goose, LibreChat, n8n, Zapier and the OpenAI Agents SDK all connect too. Choose your client above for its exact config, or see @clean-tools/mcp on npm for any stdio client.

Slim mounts — save your agent's context

Tool definitions cost context. If you only want a few tools, mount a subset with ?tools= and the server lists only those (calls to everything else still work):

Example: just the date/time tools

https://mcp.clean.tools/mcp?tools=explain_cron,validate_cron,convert_timezone,strftime_preview

REST quickstart

Every tool is also a plain HTTP endpoint at https://api.clean.tools/v1/<tool>. Here is a full request and response:

POST https://api.clean.tools/v1/convert_units

{
  "category": "length",
  "value": 26.2,
  "from": "mile",
  "to": "km"
}

200 OK

{
  "category": "length",
  "value": 26.2,
  "from": "mile",
  "to": "km",
  "result": 42.1648128
}

curl

curl -s https://api.clean.tools/v1/convert_units \
  -H 'content-type: application/json' \
  -d '{"category":"length","value":26.2,"from":"mile","to":"km"}'

JavaScript (fetch)

const res = await fetch('https://api.clean.tools/v1/convert_units', {
  method: 'POST',
  headers: { 'content-type': 'application/json' },
  body: JSON.stringify({ category: 'length', value: 26.2, from: 'mile', to: 'km' })
});
const data = await res.json();

JSON in, JSON out. Bad input returns a 4xx with `{ "error": "...", "code": "..." }`. No auth, no keys. Request bodies are capped at 128 KB. Rate limit: about 60 requests/minute per caller with burst headroom — responses carry standard RateLimit-* headers, and a 429 includes Retry-After; honor it and add jitter. Requests are processed in memory and never written to storage or logs; the only thing counted is anonymous per-tool call totals (never arguments, results, or IPs).

The tools

The same engines behind these tool pages, callable by name.

Tool	What it does	Page
`explain_cron`	Describe a cron expression and list its next run times	Open
`validate_cron`	Validate a cron expression and expand each field	Open
`expand_rrule`	Build an iCalendar RRULE and expand its occurrences	Open
`convert_timezone`	Convert a time between two time zones	Open
`strftime_preview`	Render a strftime format string for a given time	Open
`amortization_schedule`	Compute a mortgage payment and amortization schedule	Open
`tvm_solve`	Solve for any time-value-of-money variable	Open
`compound_interest`	Project compound growth with optional contributions	Open
`apr_calc`	Convert rates and solve a loan's APR	Open
`interest_rate`	Find the interest rate between two values	Open
`hash_text`	Hash text with SHA-1, SHA-256, or SHA-512	Open
`diff_text`	Line diff two blocks of text	Open
`test_regex`	Run a regular expression and return the matches	Open
`convert_data`	Convert between JSON, CSV, TSV, and XML	Open
`luhn_validate`	Validate a number with the Luhn checksum	Open
`convert_units`	Convert between units of measurement	Open
`uuid_v4`	Generate version 4 UUIDs	Open
`random_number`	Generate cryptographically random numbers	Open
`generate_password`	Generate a strong random password	Open
`decode_jwt`	Decode a JWT's header and payload and check claims	Open
`encode_decode`	Encode or decode URL and HTML text	Open
`count_text`	Count words, characters, sentences, and paragraphs	Open
`generate_qr`	Generate a QR code as an SVG	Open
`convert_case`	Convert text between cases (camelCase, snake_case, and more)	Open
`fancy_text`	Style text with Unicode glyphs (bold, italic, script)	Open
`render_markdown`	Render Markdown to sanitized HTML	Open
`lorem_ipsum`	Generate lorem ipsum placeholder text	Open
`redact_text`	Redact terms in text with block characters	Open
`format_sql`	Format and indent SQL queries	Open
`color_palette`	Build color palettes from a base color	Open
`color_contrast`	Check WCAG contrast between two colors	Open
`css_gradient`	Build a CSS gradient from color stops	Open
`percentage`	Percentage of, is what percent, and percent change	Open
`tip`	Split a bill and calculate the tip	Open
`convert_timestamp`	Convert epoch and ISO timestamps across time zones	—
`format_json`	Format, minify, and validate JSON with data-loss warnings	Open
`convert_yaml`	Convert between YAML and JSON	Open
`convert_color`	Convert a color between hex, RGB, HSL, HSV, CMYK, and OKLCH	Open
`slugify`	Turn text into URL and filename-safe slugs	—
`convert_number_base`	Convert numbers between bases 2-36 exactly	—
`query_json`	Extract values from JSON with JSONPath	—
`generate_id`	Generate UUID v5/v7, ULID, and Nano ID identifiers	Open

Schemas and discovery

OpenAPI spec Machine-readable REST schema
API catalog Well-known index of the endpoints
llms.txt Plain-text guide for language models
MCP registry: tools.clean/mcp The MCP server endpoint
npm: @clean-tools/mcp The stdio MCP package

Privacy

The API runs the same engines as the tool pages: every tool is read-only and side-effect-free, most are deterministic (same input, same output), and the generator tools use cryptographic randomness. Requests are processed in memory and never written to storage or logs, and nothing is used to train anything. The only thing counted is anonymous per-tool call totals — never arguments, results, or IPs. There are no accounts and no keys. Full details on the privacy page.