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.

How do we compare AI gateways and routers?

The compare pages score each product against the same documented capabilities and say when the review was done; the benchmark page shows measured cost and quality, including the cases the router lost.

In-house routing

OpenSmartRoute vs Building it yourself

Building internally gives you full control, and then the ongoing work: policy, outcome learning, drift detection, audit, explainability, evaluation and the maintenance after every model release. OpenSmartRoute is Apache-2.0 with zero dependencies, so you build on it rather than instead of it.

“Build on it, not instead of it.”

Choose Building it yourself when

One model, one app, nobody asks why

Choose OpenSmartRoute when

Anyone asks why a request went where it went
Compliance needs region, boundary or PII enforcement with evidence
The catalogue changes more than once a quarter
You want savings you can show finance

Capability by capability

Our reading of public documentation; corrections welcome through the support page.

Capability	OpenSmartRoute	Building it yourself
Routes to LLMs	Yes	Yes
Routes to agents, skills, tools, workflows and humansUsually a growing if/else per app	Yes	Partial
Hard policy before scoring (region, data boundary, PII, tenant, cost)Ad hoc checks; no rejection trace	Yes	Partial
Per-request trace with ranked candidates and rejections	Yes	No
Learns from your outcomes (per target, per tenant)	Yes	No
Multi-step plans (persona -> skill -> model)	Yes	No
Self-hosted, air-gapped, zero runtime dependencies	Yes	Yes
Prompt-injection and steering-gadget defences on the router	Yes	No
Output guard on responses (PII, secrets, injection: flag, redact or block)	Yes	No
Savings ledger, bill analysis and routing auditEvery request records baseline and routed cost; upload last month's provider export to see what routing would have saved before you switch	Yes	No
OpenAI-compatible proxy (chat, embeddings, Responses, moderations, audio, images)Whatever the one client library you wrap exposes	Yes	Partial
Anthropic Messages API (/v1/messages)	Yes	No
Virtual keys: per-key budgets, rate limits, allowed targets, scopes and expiryProvider keys shared by every caller	Yes	Partial
Per-target rate limits, budgets and circuit breakersRetries and a timeout, rarely a breaker or a budget	Yes	Partial
Python and TypeScript SDKs	Yes	No
MCP server for IDEs and agents	Yes	No
Browser extension and desktop app for the people on the teamPrice, personal-data and writing checks on the device before a prompt is sent, on ChatGPT, Claude, Gemini and in any desktop app	Yes	No
Provider marketplace with creditsModels by the token: every catalogue model on offer per provider with the price you pay, one key, a prepaid credits wallet and provider preferences (order, only, ignore, max_price)	Yes	No
Request/response observability UI for every callDecision trace, tokens, cost and outcome per request; prompt and answer stored when the workspace turns on request logging (redacted or full)	Yes	No

Keep your rules, add the engine

Register your existing handlers as targets with the decorator SDK and keep your rules as a RulesStrategy; the router adds policy, trace, learning and the ledger around them.

togetherpython

from opensmartroute import RouteRequest
from opensmartroute.sdk import ComponentRegistry

components = ComponentRegistry()

@components.tool("sql-report", domains=["reporting"], cost={"usd_per_1k_tokens": 0.0})
def sql_report(request, **kw):
    """Answer a reporting question with a SQL query."""  # the description strategies match against
    ...

router = components.router()  # your rules: @components.policy_rule / RulesStrategy
decision = router.route(RouteRequest(text="revenue by region last quarter"))

Prove it on your traffic

Run the Routing Audit on a week of anonymised logs: savings versus your current mix, policy violations your routing allowed, and where agents or humans would have been the better answer.

ROI calculator Try the playground

Other comparisons: OpenRouter, LiteLLM, Portkey, Cloudflare AI Gateway, Bifrost (Maxim AI), Kong AI Gateway, Helicone, Not Diamond, Martian, Unify, Requesty.

Decide on your own traffic

Run OpenSmartRoute beside Building it yourself for a week and let the trace say which one earned each request.

Scored from each product's public documentation as of 2026-10-07 - what each one documents it can do, not independent benchmark or performance results - and corrected when a vendor changes something. The number that matters is yours: send a week of anonymised logs through the routing audit, watch a live decision in the playground with every candidate ranked, and keep whichever answer holds up on your requests.

Photo: Mario Gogh

For the business

What it does for the vendor strategy.

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
Self-hosting is free
Open source, Apache-2.0
The router is a library, a sidecar or a control plane you can run yourself, so the exit is always open.
What is open

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.
How do we compare AI gateways and routers?: The compare pages score each product against the same documented capabilities and say when the review was done; the benchmark page shows measured cost and quality, including the cases the router lost.

Capability

OpenSmartRoute

Building it yourself

Routes to LLMs

Yes

Routes to agents, skills, tools, workflows and humansUsually a growing if/else per app

Yes

Partial

Hard policy before scoring (region, data boundary, PII, tenant, cost)Ad hoc checks; no rejection trace

Yes

Partial

Per-request trace with ranked candidates and rejections

Yes

Learns from your outcomes (per target, per tenant)

Yes

Multi-step plans (persona -> skill -> model)

Yes

Self-hosted, air-gapped, zero runtime dependencies

Yes

Prompt-injection and steering-gadget defences on the router

Yes

Output guard on responses (PII, secrets, injection: flag, redact or block)

Yes

Savings ledger, bill analysis and routing auditEvery request records baseline and routed cost; upload last month's provider export to see what routing would have saved before you switch

Yes

OpenAI-compatible proxy (chat, embeddings, Responses, moderations, audio, images)Whatever the one client library you wrap exposes

Yes

Partial

Anthropic Messages API (/v1/messages)

Yes

Virtual keys: per-key budgets, rate limits, allowed targets, scopes and expiryProvider keys shared by every caller

Yes

Partial

Per-target rate limits, budgets and circuit breakersRetries and a timeout, rarely a breaker or a budget

Yes

Partial

Python and TypeScript SDKs

Yes

MCP server for IDEs and agents

Yes

Browser extension and desktop app for the people on the teamPrice, personal-data and writing checks on the device before a prompt is sent, on ChatGPT, Claude, Gemini and in any desktop app

Yes

Provider marketplace with creditsModels by the token: every catalogue model on offer per provider with the price you pay, one key, a prepaid credits wallet and provider preferences (order, only, ignore, max_price)

Yes

Request/response observability UI for every callDecision trace, tokens, cost and outcome per request; prompt and answer stored when the workspace turns on request logging (redacted or full)

Yes

Keep your rules, add the engine

Register your existing handlers as targets with the decorator SDK and keep your rules as a RulesStrategy; the router adds policy, trace, learning and the ledger around them.

togetherpython

from opensmartroute import RouteRequest
from opensmartroute.sdk import ComponentRegistry

components = ComponentRegistry()

@components.tool("sql-report", domains=["reporting"], cost={"usd_per_1k_tokens": 0.0})
def sql_report(request, **kw):
    """Answer a reporting question with a SQL query."""  # the description strategies match against
    ...

router = components.router()  # your rules: @components.policy_rule / RulesStrategy
decision = router.route(RouteRequest(text="revenue by region last quarter"))

Prove it on your traffic

Run the Routing Audit on a week of anonymised logs: savings versus your current mix, policy violations your routing allowed, and where agents or humans would have been the better answer.

Run OpenSmartRoute beside Building it yourself for a week and let the trace say which one earned each request.

OpenSmartRoute vs Building it yourself

Choose Building it yourself when

Choose OpenSmartRoute when

Capability by capability

Keep your rules, add the engine

Prove it on your traffic

Run OpenSmartRoute beside Building it yourself for a week and let the trace say which one earned each request.

What it does for the vendor strategy.

No vendor lock-in

Lower cost per request

Open source, Apache-2.0

OpenSmartRoute vs Building it yourself

Choose Building it yourself when

Choose OpenSmartRoute when

Capability by capability

Keep your rules, add the engine

Prove it on your traffic

Run OpenSmartRoute beside Building it yourself for a week and let the trace say which one earned each request.

What it does for the vendor strategy.

No vendor lock-in

Lower cost per request

Open source, Apache-2.0