Task guides
15 questions answered in full, each with the tool that answers it built into the page.
Token & Cost Calculator
Estimate token usage and monthly API costs.
- How to count tokens in textEstimate tokens in any text without installing an SDK or sending it to a server. Learn why counts differ between models and where rough estimates can mislead.
- Working out Claude API cost per 1,000 requestsWork out what 1,000 Claude API requests cost. Why output tokens dominate the bill, what list pricing excludes, and how to build an estimate you can defend.
- Estimating GPT API costs before you buildBuild a defensible estimate of GPT API spend before writing code. Which variables actually move the number, and the three that teams routinely forget.
JSON Schema Builder
Build a schema visually and export it as a tool definition.
- Writing a JSON Schema for structured outputHow to write a JSON Schema that models actually follow. Why descriptions matter more than constraints, and which keywords really earn their place.
- The OpenAI function calling schema formatWhat an OpenAI function-calling tool definition looks like, how it wraps a JSON Schema, and the details that cause a definition to be rejected.
- The Anthropic tool definition formatWhat an Anthropic tool definition looks like, how input_schema differs from the OpenAI equivalent, and how to move a definition between the two.
Structured Output Validator
Check model output against a schema, with the path that broke.
- Why structured output fails validationThe five reasons model JSON fails schema validation, why each one happens, and what to change in the schema or prompt so it stops happening.
- How to validate JSON against a schemaValidate JSON against a JSON Schema and understand every error. What a property path means, which keywords matter most, and how to read a failure.
Instruction File Generator
One form, four agent instruction files that agree on every fact.
- What to put in AGENTS.mdWhat belongs in an AGENTS.md file and what does not. The sections that earn their place, and the common mistake of restating what the repo already shows.
- CLAUDE.md and AGENTS.md: which does whatHow CLAUDE.md and AGENTS.md differ, why duplicating content between the two files causes them to drift apart, and a structure that keeps both accurate.
MCP Config Validator
Catch the mistakes that break an MCP config in practice.
RAG Chunking Playground
See the chunks your settings actually produce, before you index.
- Choosing the best chunk size for RAGHow to choose a chunk size for retrieval, what the trade-off actually is, and why the right answer depends on your material rather than a rule of thumb.
- How much overlap should RAG chunks have?What chunk overlap is for, how much of it you actually need, and why going past half the chunk size makes retrieval worse rather than better.
Secret & API Key Scanner
Find leaked keys before you commit. No ads on this page.
- How to check for API keys before you commitHow to catch credentials before they reach a commit, which patterns to look for, and why a clean scan is a useful check rather than a clearance.
- I committed an API key. What now?What to do when a credential reaches a commit. Why rotation comes first, why deleting the line does not help, and the order you should work in.
How these differ from the tool pages
A tool page is organised around an instrument: what it does, how to drive it, and what its settings mean. A task page is organised around a question someone actually asked, and it embeds whichever tool answers that question directly at the top. If you already know which tool you want, the catalogue is the faster route. If you have a problem and no particular tool in mind, start here.
The distinction matters most when the answer is partly judgement. Knowing how to run a chunking preview is a tool question. Knowing how much overlap is worth paying for, and how to tell when you have too much, is a task question, and no interface can answer it on its own. These pages carry the reasoning that sits around the tool rather than inside it.
What these pages cover
Four groups, following the shape of the work. Counting and costing covers tokens, per-request pricing and building an estimate you can defend before any code exists. Schemas and structured output covers writing a schema a model will actually follow, reading a validation failure, and the two provider tool-definition formats that wrap the same JSON Schema in different clothes.
Agent instruction files cover what belongs in AGENTS.md, what does not, and how to keep a second file from contradicting the first. Configuration and credentials covers the five things that stop an MCP server connecting, how to catch a key before it reaches a commit, and what to do in what order once one already has.
Using a task page
Each page is written to be read in a few minutes and to stand on its own if the embedded tool were removed. The explanation is the substance; the tool is there so you can finish the job without moving to another page and losing your place. Nothing you type into one of these pages is uploaded, exactly as on the tool pages themselves.
Pages that address adjacent questions link to one another, so following a thread through a subject usually works better than returning here between steps. Where a page relates to a tool, the full tool page is linked at the foot of it, with the instructions, the worked example and the settings the shorter page does not repeat.