llms.txt
+116 −46
− > Free, private tools for writing better prompts for AI (now officially called SI, super intelligence, in the US). Check and fix a prompt, start from 319 prompt templates in the repository, make AI drafts sound human, check the facts in an answer, hide private details before pasting, compare models and see what AI will cost. Everything runs in the browser; prompts go straight from the visitor's browser to the AI company with their own key and never reach thinkhard.si.− - [Model Compare: compare Claude, GPT and Gemini](https://thinkhard.si/arena/): Run one prompt on up to six models side by side: Claude, GPT, Gemini and more. See speed and cost, vote blind, let an AI judge rank them. Free, no sign-up.− - [Version Compare: what changed between versions](https://thinkhard.si/diff/): Compare two versions of a prompt word by word, and see what the change does to length, token cost, structure and the score out of 10. Free, private, in your browser.− - [Fact Check List: check an AI answer before you use it](https://thinkhard.si/fact-check/): Paste an AI answer and get a checklist of every number, quote, source, link and date to verify, riskiest first, with a search link for each. Free and private.− - [Prompt Grader: score and fix any prompt](https://thinkhard.si/grader/): Paste a prompt and get a score out of 10 as you type, then answer a few plain questions to make it better. Free, no sign-up. Add a key for an AI rewrite.− - [Chatbot Safety Test: free prompt injection tester](https://thinkhard.si/injection/): See if people can trick your chatbot. Scan its instructions for weak spots, then try harmless attacks on up to three models and see exactly what leaks. Free.− - [Prompt Tester: test a prompt on real examples, free](https://thinkhard.si/lab/): Test every version of a prompt on your own examples across Claude, GPT, Gemini and more. Checks and an AI judge score each answer and show what each change broke.− - [My Prompts: save every version of your prompts](https://thinkhard.si/library/): Save prompts from any thinkhard.si tool with full version history. Compare versions, roll back and open any prompt in another tool. Stays in your browser.− - [Robot-Speak Checker: does your text sound like AI?](https://thinkhard.si/robot-speak/): Paste text an AI wrote and see what makes it sound like a robot: long dashes, filler, buzzwords, lists of three. Fix it in one click and stop it next time. Free.− - [Safe Paste: hide private details before you paste into AI](https://thinkhard.si/safe-paste/): Swap names, emails, phone, card and account numbers for placeholders before pasting into ChatGPT or Claude, then put them back in the answer. Free and private.− - [AI Cost Calculator: what your AI feature costs a month](https://thinkhard.si/tokens/): Pick what you are building and how often it runs, or set a budget, and see the monthly cost on every major AI model. Free, no API key needed.− - [MCP server: use thinkhard.si in Claude, ChatGPT and Cursor](https://thinkhard.si/developers/): Add one URL and your AI assistant can search 300+ expert prompts, grade a prompt out of 10 and price an AI feature, in Claude, ChatGPT, Cursor and VS Code.− - [AI glossary in plain English: tokens, keys, prompts](https://thinkhard.si/glossary/): Plain-English definitions of the AI words people meet at work: API key, token, system prompt, context window, prompt caching, evals, hallucination and more.− - [Learn to prompt AI well: guides, blog and use cases](https://thinkhard.si/learn/): Free guides, weekly posts, use cases by team and industry, a plain-English glossary and an AI price index, for anyone who uses AI at work.− - [AI model price index: LLM API prices compared](https://thinkhard.si/prices/): Input, output and cached-input prices for Claude, GPT, Gemini, Mistral, DeepSeek, Grok and more, per million tokens, with charts and what a typical task costs.− - [Using AI at work safely: a practical risk checklist](https://thinkhard.si/guides/ai-at-work-risk-checklist/): What AI should and should not do in a business: confidential and personal data, invented facts, approvals, prompt injection, disclosure and human review.− - [What is an API key? Why AI tools ask for yours](https://thinkhard.si/guides/api-keys/): An API key lets software use an AI model on your account. What it is, why tools ask for yours, what it costs in practice, free options and how to keep it safe.− - [How to write a prompt: brief AI like a new hire](https://thinkhard.si/guides/brief-ai-like-a-new-hire/): Write prompts the way you would brief a capable new colleague: context, goal, audience, material, format, limits and what good looks like, with office examples.− - [ChatGPT, Claude or Gemini for work: how to choose](https://thinkhard.si/guides/chatgpt-claude-gemini-for-work/): Which of ChatGPT, Claude and Gemini to use for which work task, what each offers as of October 21, 2026, and a one-hour test on your own work.− - [Choosing an AI model for business: cost, quality, speed](https://thinkhard.si/guides/choosing-an-ai-model-for-business/): Pick an AI model for business tasks in plain language: cost, quality and speed, real prices per 1,000 tasks, and the cheapest model that passes your tests.− - [Prompting Claude vs GPT vs Gemini: what differs](https://thinkhard.si/guides/claude-gpt-gemini/): What Anthropic, OpenAI and Google each recommend for prompt layout, what they share, and one prompt laid out three ways. Conventions checked October 10, 2026.− - [Extract data from invoices and PDFs with AI](https://thinkhard.si/guides/extract-data-from-documents/): Get reliable JSON or tables out of invoices, PDFs and contracts: a field sheet, schema-enforced output, blanks instead of guesses, checks in code and spot checks.− - [How to test a prompt before you ship it](https://thinkhard.si/guides/how-to-test-prompts/): Test a prompt without writing code: real test cases, exact checks, an AI judge you can trust, and a regression check each time the prompt or model changes.− - [How to write a system prompt for a chatbot or assistant](https://thinkhard.si/guides/how-to-write-a-system-prompt/): A system prompt is the standing brief a chatbot reads before every message. The seven parts it needs, a support assistant rewritten, and how to test it.− - [What an LLM call really costs: tokens and caching](https://thinkhard.si/guides/llm-cost/): How tokens are billed, the formula for input, output and cached tokens, a worked example at current list prices, and how to pick the cheapest model that passes.− - [How to measure an AI feature after launch](https://thinkhard.si/guides/measure-an-ai-feature/): What to measure once an AI feature is live: pass rate on a test set, escalations, user corrections, cost per completed task and response time.− - [Prompt engineering for beginners: the 15-minute version](https://thinkhard.si/guides/prompt-engineering-for-beginners/): Prompt engineering for people at work in fifteen minutes: the five things every prompt needs, one task rewritten, how to fix a weak answer and what to stop doing.− - [Prompt injection explained: stop system prompt leaks](https://thinkhard.si/guides/prompt-injection/): What prompt injection and system prompt leakage are, why secrets never belong in a prompt, and the layered defenses that limit the damage when an attack works.− - [AI prompts for customer support that follow policy](https://thinkhard.si/guides/prompts-for-customer-support/): A playbook for support teams using AI: replies that follow policy, triage, macros, tone and escalation, no invented refunds, and testing on real tickets first.− - [Sales and marketing prompts that don’t sound like AI](https://thinkhard.si/guides/prompts-for-sales-and-marketing/): Get outreach and marketing copy from AI that sounds like you: feed it real facts, voice samples and banned phrases, ask for variants, and check every claim.− - [How to reduce AI hallucinations at work](https://thinkhard.si/guides/reduce-ai-hallucinations/): Practical ways to reduce AI hallucinations: give the model its sources, allow “I don’t know”, ask for quotes, check citations, and test on questions with no answer.− - [Rolling out AI prompts to a team: a shared library](https://thinkhard.si/guides/rolling-out-prompts-to-a-team/): Turn good prompts into shared templates a team can trust: an owner, version numbers, review, one library, tests before every change, and when to retire one.− - [The ten checks of a well-written prompt](https://thinkhard.si/guides/ten-tests/): Task, purpose, material, scope, form, limits, standard, reference, doubt and fit: what each test asks, a weak and a fixed example, and how the score works.− - [AI prompts for support teams: replies and triage](https://thinkhard.si/use-cases/customer-support/): How support teams use AI prompts for reply drafts, ticket triage, escalations, macros and help articles, tested on real tickets and priced at your volume.− - [AI prompts for teachers and trainers](https://thinkhard.si/use-cases/education-and-training/): How schools, colleges and workplace training teams use AI prompts for quizzes, rubrics, lesson plans and family emails, with FERPA and COPPA guardrails.− - [AI prompts for finance teams: DCF, FP&A and close](https://thinkhard.si/use-cases/finance-and-accounting/): How FP&A, accounting and investment teams use AI prompts for variance commentary, the close, audit requests and DCF memos, with every number kept traceable.− - [AI prompts for HR and recruiting: hiring to policy FAQs](https://thinkhard.si/use-cases/hr-and-recruiting/): How HR and recruiting teams use AI prompts for job posts, structured interviews, feedback, onboarding and policy FAQs, with people making every decision.− - [AI prompts for legal and compliance: NDAs to DPIAs](https://thinkhard.si/use-cases/legal-and-compliance/): How in-house legal and compliance teams use AI prompts for playbook reviews, NDA triage, clause summaries and policy Q&A, with counsel making every call.− - [AI prompts for marketing teams: ads, SEO and content](https://thinkhard.si/use-cases/marketing/): How marketing teams use AI prompts for ad variants, SEO titles, landing pages, repurposed content and campaign readouts, with claim checks and real costs.− - [AI prompts for operations teams: SOPs to suppliers](https://thinkhard.si/use-cases/operations/): How operations teams use AI prompts for SOPs, supplier emails, action items, RACI charts, inbox triage and document extraction, with nothing made up.+ > Free, private tools for writing better prompts for AI, which US federal agencies have called SI, super intelligence, since a September 2026 executive order. Check and fix a prompt, start from 669 prompt templates in the repository, make AI drafts sound human, check the facts in an answer, hide private details before pasting, compare models and see what AI will cost. Everything runs in the browser; prompts go straight from the visitor's browser to the AI company with their own key and never rea+ - [Coding agent prompt checker for Claude Code and Cursor](https://thinkhard.si/agent-brief/): Paste the task you plan to give Claude Code, Codex, Cursor or Copilot. See what it is missing and what is risky, then get a better brief and AGENTS.md.+ - [Compare AI models side by side on your own prompt](https://thinkhard.si/arena/): Run one prompt on up to six models at once, from Claude and GPT to Gemini. See each answer with its speed and cost, then vote blind or let an AI judge.+ - [Custom GPT instructions generator, for Claude and Gemini too](https://thinkhard.si/assistant-builder/): Answer a few questions and get instructions for a custom GPT, Claude Project, Gemini Gem or Copilot agent, with first messages, files, settings and tests.+ - [Copy ChatGPT answers without asterisks into Word or Docs](https://thinkhard.si/copy-cleaner/): Paste a ChatGPT, Claude or Gemini answer and copy it clean into Word, Google Docs, Gmail or Slack: real headings, lists and tables, no stray asterisks. Free.+ - [Custom instructions generator for ChatGPT, Claude, Gemini](https://thinkhard.si/custom-instructions/): Answer a few questions and get custom instructions so ChatGPT, Claude, Gemini or Copilot work the way you do, with no AI filler. Free, with where to paste them.+ - [Prompt diff: compare two prompt versions word by word](https://thinkhard.si/diff/): Compare two versions of a prompt word by word and see what the change does to length, token cost, structure and the score out of 10. Free and private.+ - [AI fact checker: list every claim in an answer to verify](https://thinkhard.si/fact-check/): Paste a ChatGPT or Claude answer and get every number, quote, source and date to verify, riskiest first, with a search link for each. Free and private.+ - [Free AI prompt checker: grade any prompt out of 10](https://thinkhard.si/grader/): Paste a prompt for ChatGPT or Claude and get a score out of 10 on ten checks, with a plain question for each gap. Free and private, with no sign-up.+ - [Prompt injection tester: will your system prompt leak?](https://thinkhard.si/injection/): See if people can trick your chatbot. Scan its instructions for weak spots, then try harmless attacks on up to three models and see exactly what leaks. Free.+ - [Did the AI follow your prompt? Free instruction checker](https://thinkhard.si/instruction-check/): Paste your prompt and the AI's answer. See every rule it kept or broke, the facts it added that you never gave it, and the lines that make it hold next time.+ - [Prompt testing tool: run evals across AI models, free](https://thinkhard.si/lab/): Test every version of a prompt on your own examples across Claude, GPT and Gemini. Checks and an AI judge score each answer and show what a change broke.+ - [Prompt library audit: find duplicate and weak prompts](https://thinkhard.si/library-audit/): Paste or upload your team's prompts, or use My Prompts. See a health score, near-copies, private details, gaps and which prompts to fix first.+ - [Free prompt manager with version history, in your browser](https://thinkhard.si/library/): Save prompts from any thinkhard.si tool with full version history. Compare versions, roll back and open any prompt in another tool. Stays in your browser.+ - [Free AI prompt generator for ChatGPT, Claude and Gemini](https://thinkhard.si/prompt-builder/): Answer a few plain questions and get a well-built prompt for ChatGPT, Claude, Gemini or Copilot, scored live on ten checks. Free, no sign-up.+ - [Prompt Linter: find bugs in the LLM prompts in your code](https://thinkhard.si/prompt-lint/): Paste code or open files to find prompt injection risks, fragile JSON parsing, cache order, retired model ids and keys in prompts, with fixes and a CLI.+ - [Check if text sounds like AI: free AI writing checker](https://thinkhard.si/robot-speak/): Paste a draft and see what makes it read as AI-written: long dashes, filler, buzzwords and stock openers. Fix them in one click. Free, runs in your browser.+ - [Anonymize text before ChatGPT: free, in your browser](https://thinkhard.si/safe-paste/): Swap names, emails, phone and account numbers for placeholders before you paste into ChatGPT or Claude, then put them back in the answer. Nothing is uploaded.+ - [Token Counter for ChatGPT, Claude and Gemini: will it fit?](https://thinkhard.si/token-counter/): Count the tokens, words and pages in any text or file, see if it fits ChatGPT, Claude or Gemini, what it costs to send, and split it if too long. Free.+ - [AI cost calculator for ChatGPT, Claude and Gemini APIs](https://thinkhard.si/tokens/): Pick a task and how often it runs, and see the monthly cost on every major model, with caching and batch discounts. Free, no API key needed.+ - [Prompt chaining planner: turn a big job into AI prompts](https://thinkhard.si/workflow-planner/): Describe a job such as hiring, a launch or a board update. Get the steps in order, an expert prompt for each, and every answer carried into the next step.+ - [AI glossary in plain English: tokens, keys, prompts](https://thinkhard.si/glossary/): Plain-English definitions of nearly 100 AI words people meet at work: tokens, API keys, MCP, reasoning models, PII, hallucination and more.+ - [Learn to prompt AI: a free course, patterns and guides](https://thinkhard.si/learn/): A free library for anyone who uses AI at work: a prompt course with exercises, 24 prompt patterns, guides, fixes for bad answers, cheat sheets and a glossary.+ - [LLM API pricing comparison: price per million tokens](https://thinkhard.si/prices/): Input, output and cached-input prices per million tokens for Claude, GPT, Gemini, DeepSeek, Mistral and Grok, with the date each price was last checked.+ ## For developers: the REST API, the MCP server and the integrations+ - [Free prompt library MCP server, no sign-in needed](https://thinkhard.si/developers/): Add https://thinkhard.si/mcp to Claude, ChatGPT, Cursor or VS Code to search 669 work prompts, grade a prompt out of 10 and price an AI feature. No key.+ - [thinkhard.si API: REST, MCP, Claude API and OpenAI examples](https://thinkhard.si/developers/api/): A free REST API and MCP server: search and fetch 669 tested prompts, grade a prompt and price an AI feature from curl, Python or JavaScript. No key.+ - [MCP playground: test a public MCP server online](https://thinkhard.si/developers/playground/): Connect to a free public MCP server with no key, list its tools and prompts, call any tool and see the JSON-RPC request and answer. Copy the call as code.+ - [MCP tools reference with curl examples](https://thinkhard.si/developers/tools/): Each tool on the thinkhard.si MCP server with its parameters, the fields it returns, a tools/call request you can paste into curl and the answer it gives.+ - [thinkhard.si integrations: MCP, API, WordPress, Chrome, CI](https://thinkhard.si/integrations/): Use thinkhard.si where you already work: in Claude and ChatGPT through MCP, in WordPress, in Chrome on any page, in GitHub on every pull request, and from code.+ - [Publish Check: an AI content checker for WordPress](https://thinkhard.si/wordpress/): A free WordPress plugin that checks a post written with AI before you publish it: robotic phrasing, claims to verify, leftover markdown and private details.+ - [MCP server docs as Markdown](https://thinkhard.si/developers/mcp.md): Setup, every tool with its parameters and output fields, example calls and limits, in one Markdown file.+ - [REST API, OpenAPI 3.1](https://thinkhard.si/api/v1/openapi.json): The same tools as plain JSON at https://thinkhard.si/api/v1: search and fetch prompts, grade a prompt, price an AI feature. No key, CORS open.+ - [AGENTS.md and CLAUDE.md: what to put in them](https://thinkhard.si/guides/agents-md-claude-md/): What to put in AGENTS.md or CLAUDE.md: the commands, folders, rules and hand-back a coding agent needs on every task, what to leave out, and a full example.+ - [Using AI at work safely: a practical risk checklist](https://thinkhard.si/guides/ai-at-work-risk-checklist/): Is it safe to use ChatGPT at work? A practical checklist: what to keep out, what a person must check, approvals, prompt injection and telling people it is AI.+ - [AI prompts for Excel and Google Sheets formulas](https://thinkhard.si/guides/ai-excel-formulas/): How to get ChatGPT, Claude, Gemini or Copilot to write Excel and Google Sheets formulas that work: describe the sheet, give expected answers, test a blank row.+ - [AI for job applications: resume and cover letter, honestly](https://thinkhard.si/guides/ai-for-job-applications/): How to use ChatGPT, Claude, Gemini or Copilot on a resume and cover letter honestly: tailor your own facts, never invent, and check every line is true.+ - [Can AI grade essays? How teachers use it for feedback](https://thinkhard.si/guides/ai-grading-feedback-for-teachers/): Can ChatGPT, Claude, Gemini or Copilot grade essays? What the research shows, and how teachers use AI for rubric feedback while keeping grades and data safe.+ - [AI meeting minutes: how to write them from a transcript](https://thinkhard.si/guides/ai-meeting-minutes/): How to write meeting minutes with ChatGPT, Claude, Gemini or Copilot: a transcript, a prompt that asks for decisions and actions, and a check before you send.+ - [AI words to avoid: what makes writing sound like AI](https://thinkhard.si/guides/ai-words-to-avoid/): AI words to avoid and the habits that make writing sound like AI, from delve and seamless to filler openers, with plain swaps and a prompt that stops them.
Showing the first 40 of each.