Quick start

1. Install

npm install -g heyllm
# or: pnpm add -g heyllm

Requires Node.js ≥ 18 — heyllm uses the built-in global fetch, so there’s no extra HTTP dependency to pull in.

2. Set your API key

heyllm reads your key from the OPENAI_API_KEY environment variable. Export it once:

export OPENAI_API_KEY="sk-..."

If no key is found and --api-key isn’t passed, heyllm prints a clear error to stderr and exits with code 2. See Configuration for per-call overrides and other providers.

3. Ask your first question

heyllm "explain monads simply"

A monad is a way to wrap a value along with a recipe for chaining
operations on it without unwrapping by hand...

Tokens stream to your terminal as the model produces them. That’s the whole single-shot flow — no session, no history, no config file.

4. Pipe stdin in as context

The real power is piping. The positional prompt becomes the instruction, and whatever you pipe in becomes the context — combined into one user message, prompt first:

git diff | heyllm "write a conventional commit message"

feat(parser): handle SSE frames split across network chunks

Any command that writes to stdout works as context:

cat error.log | heyllm "what's the root cause here?"
curl -s https://example.com | heyllm "summarize this page in 3 bullets"

You can also pipe input with no prompt at all and let stdin stand on its own:

echo "translate to French: good morning" | heyllm

Where to next

• CLI reference — every flag, streaming vs --no-stream, --json, and exit codes.
• Configuration — point --base-url at OpenRouter, a local Ollama, or any OpenAI-compatible API, and pick a model.
• Recipes — a handful of genuinely useful one-liners.