Can we run it inside our own network?

Yes. The router runs as a container or a Helm release on your infrastructure, and a routing rule keeps regulated data on the models you allow.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

How much can we save on our LLM bill?

On the public benchmark the router saved 89% against sending everything to the reference model. Import last month's provider bill on the bill page and the estimate is computed on your own traffic before you change a line of code.

Software engineering · use case

The right model for each coding task, from a one-line fix to a redesign.

Route coding requests by what they need: completions and small fixes to a fast, cheap model; refactors and reviews to a coding agent; architecture and security questions to a reasoning model. Works through the MCP server in VS Code, Cursor and Claude, with the cost of each request shown.

Try OpenSmartRoute free Talk to sales about a pilot

Free plan, no card. Self-hosting is free forever; the rules below run the same way on both.

The situation

What the team is dealing with today

An IDE assistant uses whatever model it was configured with for everything - a variable rename and a cross-module refactor cost the same and get the same attention. Switching is a setting a person flips, and nobody knows what the day's completions cost until the invoice.

What changes

Small edits and completions on a fast, cheap model; real code on the coding agent
Prose, commit messages and documentation kept off coding-model prices
A quote before an expensive request and the model named next to every answer

One request

What happens, step by step

Code requests go to the best coding model; everything else stops paying coding-model prices.

Request arrives
Code and complexity detected
Prose is kept off the coding agent
Coding agent answers the real code
Cost and alternatives on the trace

Route a request like this in the playground and read the trace

A software engineering request, quoted live“Refactor this Python function to stream the file instead of reading it whole, keep the public signature, and add a test for an empty file.”

Quoting every candidate in the public catalogue...

The policy

Two text files, not a rewrite

This is the rules.yaml that produces the behaviour above; targets.yaml names the models, agents and people it refers to. Complexity is a signal, so a one-line regex question never wakes the coding agent.

rules.yaml - software engineering (reference configuration)yaml

rules:
  - name: real-code-to-the-coding-agent
    when:
      domains: [coding]
      min_complexity: 0.45
    prefer: [coding-agent]
    weight: 0.7

  - name: prose-never-pays-coding-prices
    when:
      domains: [general_chat, marketing, hr]
    avoid: [coding-agent]
    weight: 0.6

Example: Illustrative configuration or synthetic data, not a customer's.A reference configuration to start from - rules the router enforces as written, not a description of how a named customer runs it.

Other use cases

For the business

What it does for the whole organisation.

Cloud, self-hosted or on-premises
Run it where the data is
The same router ships as a hosted platform, a container image and a Helm chart; the data boundary is a routing rule.
Deployment options
One URL, every vendor
No vendor lock-in
Point your OpenAI-compatible client at one address; add, swap or retire providers in the catalogue without touching application code.
Every vendor side by side
89% less
Lower cost per request
On the 150-prompt benchmark the router spent 89% less than sending every prompt to the reference model, and the judged quality went up, not down.
Read the benchmark

Can we run it inside our own network?: Yes. The router runs as a container or a Helm release on your infrastructure, and a routing rule keeps regulated data on the models you allow.
Will this lock us into another vendor?: No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.
How much can we save on our LLM bill?: On the public benchmark the router saved 89% against sending everything to the reference model. Import last month's provider bill on the bill page and the estimate is computed on your own traffic before you change a line of code.

Start with the software engineering rules and change what you disagree with.

OpenSmartRoute is open source and the free plan keeps the full trace of every decision. Create a workspace, paste the rules, route a request.

Create a free account Read the benchmark

Free plan, no card. Fifteen thousand decisions a month with the full trace.

rules: - name: real-code-to-the-coding-agent when: domains: [coding] min_complexity: 0.45 prefer: [coding-agent] weight: 0.7 - name: prose-never-pays-coding-prices when: domains: [general_chat, marketing, hr] avoid: [coding-agent] weight: 0.6

The right model for each coding task, from a one-line fix to a redesign.

What the team is dealing with today

What happens, step by step

Two text files, not a rewrite

What it does for the whole organisation.

Run it where the data is

No vendor lock-in

Lower cost per request

Start with the software engineering rules and change what you disagree with.

The right model for each coding task, from a one-line fix to a redesign.

What the team is dealing with today

What happens, step by step

Two text files, not a rewrite

What it does for the whole organisation.

Run it where the data is

No vendor lock-in

Lower cost per request

Start with the software engineering rules and change what you disagree with.