# Top AI CLIs — How To Feed Them The Prompts Our Generators Build Claude Code, Gemini CLI, aichat, OpenAI CLI, llm, and continue. Seven command-line AI tools and a concrete pattern for feeding each one the mega-prompts our site builders and analyzers generate. Author: J.A. Watte Published: April 20, 2026 Source: https://jwatte.com/blog/blog-ai-clis-with-our-prompts/ --- Every generator on jwatte.com outputs a prompt. [Single Site Gen](/tools/single-site-gen/) writes a multi-thousand-word site build prompt. The [Mega Analyzer](/tools/mega-analyzer/) emits a prompt that covers every failed signal. The [Syndication Planner](/tools/syndication-planner/) writes a per-platform rewrite brief. All of them are designed to be pasted somewhere. The question most readers don't ask out loud: *pasted where, exactly.* ChatGPT and Claude.ai are the obvious answers. But for anyone working on the command line. Especially when you're iterating on a prompt, feeding it different context, or piping files in. A CLI is a better fit. Here's the short list. ## Claude Code Anthropic's own CLI. Installs via `npm install -g @anthropic-ai/claude-code`. You run `claude` in a project directory and it opens an interactive session inside your terminal with full file read/write access. Usage pattern with our generators: copy the generated mega-prompt to your clipboard, run `claude`, paste. Claude gets the prompt and your project files in one session and can make the edits directly. It's the most capable for multi-file work. Agentic by default. It can plan, read, edit, test, commit. For anything more complex than a single-file rewrite, this is the one. ## Gemini CLI `npm install -g @google/gemini-cli`. Similar shape to Claude Code. An agentic session inside your terminal. Models: Gemini 2.5 Pro / Flash / Ultra. Connects to Google Search built-in, which is useful when the prompt asks for citation-ready research. Usage pattern: `gemini` to start. Then paste. Works well for research-heavy prompts (the SERP Features, Entity Citation Radar outputs) where web-grounded answers matter. ## aichat A one-off Rust CLI from sigoden. `cargo install aichat` or grab a release binary. Supports most model providers (OpenAI, Anthropic, Google, Ollama, Mistral). Shines for one-shot queries and pipe-in workflows. Usage pattern: copy the prompt to a file (`prompt.md`), then `aichat --file prompt.md`. Or pipe straight in: `cat prompt.md | aichat`. For the Single Site Gen output, `aichat --file build-prompt.md > build-output.md` writes the entire generated site files straight to disk. ## OpenAI CLI (and `llm`) Simon Willison's `llm` tool (`pip install llm`, then `llm keys set openai`) is the Swiss Army knife. Supports providers via plugins. OpenAI, Anthropic, Gemini, Mistral, local Ollama, anything with an API. Usage pattern: `llm -m claude-opus-4-7 < prompt.md`. Or pipe from clipboard on macOS: `pbpaste | llm -m gpt-5`. Logs every conversation to a local SQLite database. Easy to search your prompt history later. For the Mega Analyzer's fix prompts specifically, `llm` pairs well because you can run the same prompt through three different models and diff the outputs: `cat fix-prompt.md | llm -m claude-opus-4-7 > a.md && cat fix-prompt.md | llm -m gpt-5 > b.md && diff a.md b.md`. ## Continue (VS Code / JetBrains) Not a terminal CLI, but worth mentioning because it's the IDE-side equivalent. Install the Continue extension, hit `Cmd+L`, paste the prompt. Continue's project context picker lets you @mention specific files so the prompt has the right working set. Usage pattern: generate a prompt from our tools, open your project in VS Code, paste into Continue's chat with the relevant files attached. For site-build work where the LLM needs to know what's already in the repo, this is better than a blind Claude Code session. ## Aider `pip install aider-chat`. Like Claude Code but git-native: every edit is a commit. For people who want the LLM's changes on an inspectable commit history from the start. ## The pattern that actually ships The generators on this site produce plain text. Text plays everywhere. Our usage pattern, roughly: - **Single-URL audits** (Mega Analyzer, Site Analyzer, E-E-A-T, Discover Readiness): paste into Claude.ai or ChatGPT web. Fastest. - **Site builds** (Single Site Gen): download the `.md`, feed it to `claude` in your project directory, let it write files. - **Batch audits** (Mega Batch, Batch Compare): run through `llm` so you can re-run the same prompt against different models. - **Research-heavy tools** (SERP Features, Entity Citation Radar, FAQ Harvester): Gemini CLI, because grounded search matters. - **Multi-tool workflows** (Syndication Planner output + Newsletter Swap brief + content calendar): `aichat` with file chaining. The generated prompts are format-agnostic. Whichever of these fits your stack. Use it. And before you upgrade to a new model version in any of these tools, run through the [pre-upgrade validation checklist](/blog/blog-validate-ai-model-before-upgrade/) first. Model updates have broken production workflows more than once. ## How to get the most out of a mega-prompt Three habits pay back across all tools. **Pick the right level.** The Mega / Quick / Mini toggle exists because Mega prompts are genuinely better for full builds and genuinely overkill for a single-file fix. Matching prompt size to task size is free iteration speed. **Check skill-level preamble.** Site Analyzer and Mega Analyzer let you pick Beginner / Intermediate / Advanced. Advanced skips explanations and ships code; Beginner explains every fix first. If the model is producing more prose than you want, bump the level. **Run the audit after.** Every site-build prompt is paired with an audit that grades the result. Single Site Gen → [Mega Analyzer](/tools/mega-analyzer/). Speakable Gen → [AI Citation Readiness](/tools/ai-citation-readiness/). The build-prompt-then-audit-prompt loop is the whole reason these tools exist together. --- Canonical HTML: https://jwatte.com/blog/blog-ai-clis-with-our-prompts/ RSS: https://jwatte.com/feed.xml JSON Feed: https://jwatte.com/feed.json Hero image: https://jwatte.com/images/blog-ai-clis-with-our-prompts.webp