About quickhelp.dev

What is quickhelp.dev?

quickhelp.dev is a single-domain collection of small, deterministic developer tools. Every tool is free, stateless, and completes in under 5 seconds. Each one exposes four interfaces: a human UI at /{slug}, a REST API at /api/{slug}, an OpenAPI 3.1 schema at /openapi.json, and an MCP server entry at /mcp.

The site currently hosts 15 tools across encoding, formatting, conversion, cryptography, text processing, and more. Each tool comes with a detailed guide explaining what it does, how to use it, and when to use it — not just a form to fill in. New tools are added regularly, chosen based on search demand and implementation feasibility.

Why we built this

The immediate trigger was spending too much time Google-searching for basic conversions during development — decoding a JWT to check its expiry, formatting a JSON blob to read a nested field, converting a Unix timestamp to a readable date. The existing tools are scattered across dozens of sites, many of which are cluttered with ads, require accounts, or silently send your data to a server. A JWT containing real credentials should not need to leave your browser to be decoded.

The longer-term motivation is the shift toward AI-agent workflows. When an agent calls an external tool, it needs to be able to trust that tool — that the same input always produces the same output, that the interface is documented in a machine-readable format, and that the tool does not have hidden side effects. quickhelp.dev is designed for that trust model from the start: strict input and output schemas, no server-side state, and a single OpenAPI document that lets any agent or automation framework discover and call every tool without manual configuration.

Design principles

• One tool, one job. Every tool does exactly one thing. A JWT decoder decodes JWTs. A JSON formatter formats JSON. A hash generator generates hashes. Narrow scope means a predictable, trustworthy interface.
• No auth, no database. Every tool is stateless. Nothing is stored server-side beyond what is necessary to return the response. There are no user accounts because there is nothing to persist.
• Deterministic output. Given the same input, every tool returns the same output. No randomness, no AI-generated results, no variation based on time or server state (except the UUID generator, where randomness is the feature).
• Privacy-first for sensitive data. Image tools (converter, resizer, background remover) run entirely in your browser using the Canvas API and WebAssembly. Your files never leave your device.
• Agent-native from day one. Every tool is discoverable by AI agents via /openapi.json, /llms.txt, and the MCP endpoint. The schema definitions are designed to be unambiguous, so agents can call tools correctly on the first attempt without trial and error.
• Free tier with attribution. The API is free to use. Anonymous API responses include a small attribution watermark. Paid keys (coming soon) will remove the watermark and include higher rate limits.

Technical architecture

The site runs on Next.js 14 with the App Router, deployed on Vercel. Tool logic lives in independent packages (packages/tools/<slug>/), each exporting a manifest that includes the tool's name, description, category, input schema (Zod), output schema (Zod), handler function, and SEO content block.

The four discovery surfaces — web UI, REST API, OpenAPI spec, and MCP server — are all generated automatically from the same registry. Adding a new tool means writing one file and adding it to the registry; everything else (routing, API endpoint, schema generation, MCP registration, sitemap, llms.txt) is handled automatically. This architecture makes it practical to add multiple tools per week without accumulating maintenance debt.

For AI agents

quickhelp.dev exposes the following machine-readable discovery surfaces:

• /openapi.json— OpenAPI 3.1 schema for all 15 tools, with full input/output schema definitions
• /llms.txt— llms.txt discovery document (compact format, one tool per section)
• /llms-full.txt— Full content dump including schemas and examples — designed for one-shot context loading
• /mcp— Model Context Protocol endpoint (HTTP transport) — compatible with Claude, Cursor, and any MCP client

Contact

Built by Jan Stepien. Questions, bug reports, or feature suggestions: [email protected]. Alternatively, visit the Contact page for more details on response times and what to include.