Map

routing.run

Toolbox toolboxllmapi-routerwatchlistTypeScriptopen source (license unspecified at time of ingest) โ†ณ show in map Markdown
title
routing.run
type
toolbox
summary
Model-routing layer with one OpenAI-compatible endpoint and flat per-request pricing
tags
llm, api-router, watchlist
language
TypeScript
license
open source (license unspecified at time of ingest)
created
2026-05-10
updated
2026-05-10

routing.run is a model-routing layer that sits between an application and multiple LLM providers, exposing one OpenAI-compatible endpoint with fallback chains, live routing control, and per-request pricing instead of per-token billing. The pitch: "providers change, your integration doesn't."

The differentiator vs OpenRouter and Chutes is the pricing model. OpenRouter and Chutes charge per million tokens, so cost grows with prompt and completion length. routing.run charges per request โ€” flat regardless of context size, which gets attractive once you're sending 32kโ€“128k token prompts (the chart on the homepage shows the per-token model crossing past $3k/month for 128k-token avg requests vs flat lines for routing.run). Predictability over absolute minimum cost, basically.

Stated guarantees:

  • Open source routing infrastructure (so users can audit how requests are handled)
  • Zero prompt logging (no recording of prompts or completions)
  • No IP tracking
  • Smart provider failover โ€” anomalous-latency and outage detection routes traffic to healthy fallbacks

How it would slot in: drop-in replacement for OPENAI_BASE_URL, configure provider preference and fallback order in their dashboard, your application code stays the same.

Limitations / open questions

  • License unspecified on the homepage โ€” "fully open source" is claimed but the repo isn't linked from the marketing page; check before adopting
  • Founded recently; no track record yet
  • Per-request pricing only beats per-token pricing once your average context size is large; small-context use cases still win on per-token providers
  • The "zero logging" claim is a contractual commitment; verify in their privacy policy and (if you can) via traffic analysis on a test deployment

Watchlist reason

On watchlist: tiny new project, no track record, closed-source-by-default-but-claimed-open-source ambiguity. Re-check in 90 days for visible repo, license clarity, customer references, and whether the per-request economics hold up against OpenRouter's evolving pricing.

Cross-references: llm-api-routing-layer for the broader pattern.