llm-routing

Stop hitting the limit on your Claude Pro or Max plan.

Routes the routine prompts — "what does this error mean", "reformat this JSON", "is the service up" — to free and cheap models, so your subscription quota is still there when you need it at 4pm. Same workflow, same commands. The model choice changes underneath.

npx -y llm-routing install

No API keys required. No hosted proxy. No account.

Why this package is 3 kB

It contains three files, and that is deliberate.

llm-router is a Python program shipped as a standalone binary — roughly 300 MB per platform, because it carries its own interpreter and a large provider library. Vendoring all four platforms into one npm tarball would exceed a gigabyte and hand every user the three they cannot run.

So postinstall downloads the one binary that matches your platform from the matching GitHub release, and bin/llm-router.js hands off to it. That is the whole package.

If the download fails, the install fails loudly rather than leaving a broken llm-router on your PATH to be discovered mid-session.

Why a proxy cannot do this

Every other router in this category is a proxy: you point your agent at a local endpoint and it forwards requests using your API keys. That design has a hard limit — a proxy cannot intercept a session authenticated by a subscription, because there is no key to forward.

If you pay per token, a proxy serves you well and there are good ones. If you pay a flat monthly fee and the thing you run out of is quota, this is the gap it fills.

Supported platforms

| Platform | Status | |---|---| | macOS arm64 | supported | | macOS x86_64 | supported | | Linux x86_64 | supported | | Windows x86_64 | supported |

On an unsupported platform the install refuses cleanly and points you at pip install llm-routing, which works anywhere Python 3.11+ does.

macOS: Gatekeeper

The binaries are not yet notarised, so macOS quarantines them on first run. If llm-router will not start:

xattr -dr com.apple.quarantine "$(dirname "$(readlink -f "$(which llm-router)")")"

After installing

llm-router install     # wire up your host (Claude Code by default)
llm-router doctor      # verify everything is connected
llm-router status      # today's savings and quota

Links

• Source, docs and issues — https://github.com/ypollak2/llm-router
• PyPI (pip install llm-routing) — https://pypi.org/project/llm-routing/
• Benchmark write-up, including what did not work — docs/ROUTERARENA.md
• Host support matrix, including what each host genuinely cannot do — guide/HOST_SUPPORT_MATRIX.md

MIT licensed.