Guides for building with LLMs
Practical, sourced guides on personal AI agents (ChatGPT Work, Claude Code, Claude Duet, Comet, Manus, Doubao, Qwen, Kimi), getting LLM API keys, tokens and API cost, context windows, exporting your AI chat history, MCP servers, Structured Outputs, and writing AGENTS.md, CLAUDE.md, SKILL.md and Cursor rules. Every vendor claim links to official documentation; each guide shows the date it was checked.
TL;DR: Short, sourced guides on personal AI agents, LLM API keys, token pricing and cost, context windows, chat export, MCP, Structured Outputs and agent instruction files — every vendor claim links to official documentation.
Recommendations & comparisons
Cheapest LLM APIs (October 2026): Best-Value Models for Low Cost, Coding, Long Context and Chinese Vendors
“Cheapest” depends on how you use a model: a chatbot, a coding agent that re-sends the same files with caching, and a 500K-token document job rank models differently. Every number on this page is computed from the same official price table our token counter and cost calculator use (checked 2026-10-22), so the rankings move when vendors change prices.
Read guide →
Best Personal AI Agents in 2026: Top 3 Picks for Office Work, Coding, Browser Tasks, Free Trials and China
There is no single best AI agent; the right one is the one that already holds your context — your files, your code, your browser or your Chinese apps. Below are our top three for five common situations, based only on what each vendor’s official page says (the same facts as our full comparison, checked 2026-10-21).
Read guide →
Free Token Counters and LLM Cost Calculators Compared: og.cc.cd, OpenAI Tokenizer, Tiktokenizer, tiktoken and Vendor APIs
Free token counters answer different questions. Some show exactly how OpenAI’s tokenizer splits text, some build a full chat payload, and the vendors’ own APIs give billing-grade counts for their models. Here is what each common free tool does, from its own page, and where our tools fit.
Read guide →
Tokens & API cost
How LLM Tokens Work — and How to Estimate API Cost Before You Ship
Every LLM API bills by the token, but a token is not a word or a character. This guide explains what tokens are, why the same text produces different counts on different models, and how to turn a token count into a monthly budget you can defend.
Read guide →
LLM API Price Comparison: OpenAI, Claude, Gemini, Grok, DeepSeek, Qwen, Kimi, GLM and More (October 2026)
Two tables, 32 models, 13 vendors — every number copied by hand from the vendor’s official pricing page, with a link to the source. The first table (OpenAI, Anthropic, Google, DeepSeek, Qwen) was checked on 2026-10-18; the second (Grok, Mistral, Kimi, GLM, Doubao Seed, Hunyuan, MiniMax, MiMo) was added on 2026-10-21. Below them: the fine print that changes what you actually pay.
Read guide →
Reduce LLM Token Usage: Prompt Caching Basics for OpenAI, Claude, Gemini and DeepSeek
Most LLM spend is repeated input: the same system prompt, tools and documents sent again on every call. This guide covers the cheap wins first, then how prompt caching works on each major API and how to structure a prompt so the cache actually hits.
Read guide →
Exporting chat history
How to Export Your Claude Chat History (and Read the conversations.json File)
Claude can email you a complete copy of your conversations. This guide walks through the official export, explains what is inside the file, and shows how to turn it into something you can actually read, search and archive.
Read guide →
How to Export ChatGPT Conversations and Read conversations.json
ChatGPT’s data export gives you a ZIP with every conversation in a file called conversations.json. The file is not a simple list of messages — it is a tree. This guide shows how to request the export and how to read that tree correctly.
Read guide →
How to Save and Export Gemini Chats: Docs, Share Links, Takeout and Markdown
Unlike Claude and ChatGPT, the Gemini app does not hand you a single file of every conversation. You can still keep your chats — you just need to pick the right tool for the job. Here are the official options and their trade-offs.
Read guide →
Instruction files for coding agents
What Is AGENTS.md? How to Write One That Coding Agents Actually Follow
AGENTS.md is a plain Markdown file that tells AI coding agents how to work in your repository: how to build and test, which conventions to follow, and what not to touch. Think of it as a README written for agents instead of humans.
Read guide →
CLAUDE.md Best Practices: Writing Project Memory for Claude Code
CLAUDE.md is the file Claude Code reads at the start of every session to learn your project’s commands, conventions and rules. Done well, it saves you repeating yourself; done badly, it wastes context and gets ignored. These practices come straight from Anthropic’s documentation.
Read guide →
How to Write a SKILL.md: Agent Skills Format, Fields and Examples
An Agent Skill is a folder with a SKILL.md file that teaches an agent how to do one specific job — fill a PDF form, run a release, follow your API conventions. The agent sees only the short description until the skill is needed, so skills can hold detail that would be too expensive to keep in every prompt.
Read guide →
Cursor Rules Guide: .cursor/rules, .mdc Files and How They Relate to AGENTS.md
Cursor gives you two ways to tell its agent how your project works: structured rules in .cursor/rules, and a plain AGENTS.md. This guide explains the rule format from Cursor’s documentation, when each kind of rule loads, and how to combine rules with AGENTS.md without repeating yourself.
Read guide →
MCP: connecting agents to tools
What Is MCP (Model Context Protocol)? A Beginner’s Explainer
MCP is the open standard that lets AI apps such as Claude, ChatGPT, VS Code and Cursor plug into outside tools and data. This guide explains what it is, how the pieces fit together, and what actually happens when you “add an MCP server”.
Read guide →
How to Set Up MCP Servers in Claude Desktop, Cursor and VS Code
The three most popular MCP hosts all read a JSON file, but the file names, locations and top-level keys differ. Here is exactly what each one expects, checked against the official docs, plus the fixes for the problems people hit most.
Read guide →
Structured outputs, context & cost planning
OpenAI Structured Outputs: Writing a JSON Schema That Strict Mode Accepts
Structured Outputs makes an OpenAI model’s reply match your JSON Schema instead of merely being valid JSON. The catch is that strict mode accepts only a subset of JSON Schema. These are the rules, where to put the schema, and how to handle the cases where the model still does not return it.
Read guide →
LLM Context Window Comparison: GPT-6, Claude 5.5, Gemini 4, DeepSeek and Qwen
Most flagship models now advertise a context of about a million tokens, but “context window” hides three different limits and, for some models, a price jump. Here are the current numbers from each vendor’s own model pages and what they mean in practice.
Read guide →
How to Estimate LLM API Costs: A Method With Worked Examples
A defensible LLM budget comes from four numbers you can measure — requests, input tokens, output tokens and cache hit rate — multiplied by prices you can cite. This guide walks through the method and three worked examples using list prices checked on 2026-10-20.
Read guide →
Model migrations
API keys
How to Get an LLM API Key: OpenAI, Claude, Gemini, DeepSeek, Qwen, Kimi, GLM and More
Every LLM API starts with the same chore: find the console, create a key, work out billing and region rules, and make a first call. This page compares ten providers side by side, using only what their official documentation says (checked 2026-10-20), and links to a step-by-step guide for each.
Read guide →
How to Get an OpenAI API Key: Steps, Region Rules, Billing and First Request
OpenAI API keys are created in the API Dashboard at platform.openai.com. This guide covers the exact pages, who can sign up, how billing and usage tiers work, and a first request you can paste into a terminal.
Read guide →
How to Get a Claude API Key in the Claude Console (Anthropic)
Claude API keys are created in the Claude Console at platform.claude.com. The key is shown only once, so read the steps before you click. This guide also covers supported regions, usage tiers and the headers every request needs.
Read guide →
How to Get a Gemini API Key in Google AI Studio (Free Tier, Billing, Key Types)
Gemini API keys come from Google AI Studio. New users get a default project and key automatically; existing Google Cloud users import a project first. Google changed the key type in 2026, and the free tier has conditions worth knowing before you build on it.
Read guide →
How to Get a DeepSeek API Key: Console, Balance, Base URL and Rate Limits
DeepSeek’s API keys live on its open platform at platform.deepseek.com. The API speaks both the OpenAI and the Anthropic format, so most SDKs only need a new base URL and key.
Read guide →
How to Get an Alibaba Cloud Model Studio (Bailian) API Key — China & Singapore
Alibaba Cloud Model Studio (阿里云百炼) serves Qwen and other models through an OpenAI-compatible API. Keys, domains and models are separate per region, so decide first whether you need the China (Beijing) site or the international (Singapore) site.
Read guide →
How to Get a Xiaomi MiMo API Key: Xiaomi Account, Real-Name Rules and Base URLs
Xiaomi MiMo’s API platform (platform.xiaomimimo.com) uses your Xiaomi account. Domestic and overseas accounts are billed and served separately, and pay-as-you-go keys differ from Token Plan keys — this guide sorts out which key and which base URL you need.
Read guide →
How to Get a Kimi (Moonshot AI) API Key: Verification, Voucher, Top-Up and Base URL
Moonshot AI’s Kimi API is managed on platform.kimi.com (the platform.moonshot.cn docs now redirect there). The key itself is quick to create; what trips people up is verification, the voucher’s limits, and rate limits that depend on how much you have topped up.
Read guide →
How to Get a Zhipu GLM API Key on bigmodel.cn or Z.ai
Zhipu runs two developer platforms for its GLM models: open.bigmodel.cn in China and z.ai internationally. Both use an OpenAI-style API with Bearer auth; the main trap is that plan keys and standard API keys are not interchangeable.
Read guide →
How to Get a Volcengine Ark (Doubao) API Key: Create, Enable Models, First Call
Volcengine Ark (火山方舟) is ByteDance’s model platform for Doubao and other models. Getting a key takes a minute, but a key alone is not enough — each model has to be enabled before you can call it.
Read guide →
How to Get an OpenRouter API Key: Credits, Payment (incl. AliPay), Free Models
OpenRouter puts many providers’ models behind one OpenAI-compatible API and one key. You pay in US-dollar credits, so the order is simple: create an account, add credits, then create a key.
Read guide →
Personal AI assistants & agents
Personal AI Agents Compared: ChatGPT, Claude, Gemini, Copilot, Perplexity Comet, Manus and China’s Doubao, Qwen, Kimi, GLM (October 2026)
“AI agent” now means an assistant that does the work instead of only answering: it plans a task, opens a browser or your files, runs for minutes or on a schedule, and hands back a document, a site or a finished form. Ten products sell this to individuals. This page compares what each one says it can do, what the paid tiers cost, and who it suits — using only the vendors’ own pages, checked on 21 October 2026.
Read guide →
ChatGPT Work: How to Turn It On and Run Your First Agent Task
ChatGPT Work is how ChatGPT acts across your apps, files and browser to produce finished documents, decks, spreadsheets and sites. Here is who gets it, what it costs, and how to run a first task without handing over more access than you need.
Read guide →
Claude Code: Install, Sign In and Run Your First Coding Task
Claude Code is Anthropic’s agentic coding tool: it reads your codebase, edits files, runs commands and works with git. It runs in the terminal, VS Code and JetBrains, a desktop app and the browser. This guide uses Anthropic’s own install commands and plan table.
Read guide →
Perplexity Comet: Install the AI Browser and Use Its Browser Agent
Comet is a web browser with Perplexity built in. The browser itself is free; its browser agent, Comet Assistant, can work through multi-step web tasks for you, and that part depends on your Perplexity plan.
Read guide →
Manus AI Agent: Free Credits, Chat vs Agent Mode, and Which Plan to Pick
Manus is a general-purpose agent that runs in the cloud: you describe an outcome, it researches, writes, builds slides or a website and hands back the result. It is metered in credits, and the free plan is generous enough to test it properly.
Read guide →
Claude Duet: What Anthropic’s Proactive Personal Agent Does, Who Can Get It, and How to Join
On 21 October 2026 Anthropic opened a research preview of Claude Duet, a separate app in which Claude works as a proactive agent: it tells you what needs your attention, takes tasks you message it, keeps working when you step away and checks with you before acting. This page sums up what Anthropic’s own pages say.
Read guide →