Switchyard

Switchyard routes each LLM call to the cheapest model that can still do the job. Without changing a line of your agent.

Get started →

*Total cost based on average ISP token cost

What is Switchyard

Switchyard picks which model serves each LLM call.

Use Switchyard

Switchyard runs inside gateways you may already have.

• NeMo Relay — a native plugin. Load a routes.toml into a Relay deployment you already run. Setup →
• LiteLLM — a routing plugin for LiteLLM's Router and proxy. examples/litellm
• More integrations coming soon.

flowchart LR
    subgraph R["LiteLLM · NeMo Relay"]
        P["Switchyard"]
    end
    P--> M["Efficient model"]
    P--> N["Capable model"]
    P--> O[etc.]
    G[You] -->|"request"| P
    style P fill:#76B900,stroke:#5A8F00,color:#000

Loading

Integrate Switchyard into your gateway or harness

Embed the routing algorithms in your own. Switchyard picks the model; your harness makes the call, so your transport, retries, and credentials stay untouched.

• Install: pip install nemo-switchyard
• Then follow Path 2 — Embed the Library: construct an algorithm, drive its step stream, make the answer call.
• Also available for Rust as switchyard-libsy; Path 2 has the Cargo.toml block.

flowchart LR
    subgraph R["Your LLM gateway / harness"]
        P["Switchyard"]
    end
    P--> M["Efficient model"]
    P--> N["Capable model"]
    P--> O[etc.]
    G["Your users"] -->|"request"| P
    style P fill:#76B900,stroke:#5A8F00,color:#000

Loading

Run Switchyard as a standalone proxy

A server in front of an agent, when you have no gateway to put Switchyard in:

cargo install --locked switchyard-server
switchyard-server --config routes.toml --port 4000

Point Claude Code, Codex CLI, or any OpenAI/Anthropic SDK client at the proxy. Switchyard decides per turn which model serves it.

flowchart LR
    P["Switchyard<br/>standalone proxy"]
    P--> M["Efficient model"]
    P--> N["Capable model"]
    P--> O[etc.]
    G[You] -->|"unchanged native API"| P
    style P fill:#76B900,stroke:#5A8F00,color:#000

Loading

Components

Pre-1.0 software. APIs, configuration, and routing behavior can change between releases — pin the version you integrate.

Component	Stability	Use it for	Guidance
switchyard-libsy	Beta	Routing embedded in your own gateway or harness. You own model calls, credentials, and retries.	Trial integrations. API will change before v1.0.
switchyard-llm-client	Alpha	HTTP model calls and protocol translation alongside libsy.	Experiments and pilots.
switchyard-runner	Alpha	Running configured routes inside another runtime, such as NeMo Relay.	Integration work and supervised pilots.
switchyard-server	Demo	A standalone OpenAI- and Anthropic-compatible proxy.	Demos and evaluation only. Not for production.

Get Started

Three paths, in the same order as above. Each is self-contained: start at step 1, stop when you reach the result named under the heading.

Path 1 — Load the NeMo Relay Plugin

You finish with an existing NeMo Relay deployment routing through Switchyard. Requires NeMo Relay >=0.8.1,<0.9.0.

1. Build the plugin bundle.

python crates/switchyard-nemo-relay-plugin/scripts/package_bundle.py

2. Write the Switchyard deployment to /etc/switchyard/routes.toml — the same version-1 TOML the proxy uses. Copy the file from step 2 of Path 3 below.

3. Point Relay at the generated manifest. Use exactly one deployment source: a path, as here, or the config nested under switchyard_config.

[[plugins.dynamic]]
manifest = "./plugins/switchyard/relay-plugin.toml"

[plugins.dynamic.config]
priority = 0
switchyard_config_path = "/etc/switchyard/routes.toml"

4. Restart Relay. It now runs any algorithm switchyard-runner supports, while Switchyard owns provider HTTP dispatch.

Details: switchyard-nemo-relay-plugin and the server configuration guide.

Path 2 — Embed the Library

You finish with your own harness picking a model per request and still making every model call itself. Shown in Python; the Rust API has the same shape.

1. Install.

pip install nemo-switchyard

For Rust, add the crates to your Cargo.toml instead:

[dependencies]
async-trait = "0.1"
futures = "0.3"
switchyard-libsy = { git = "https://github.com/NVIDIA-NeMo/Switchyard.git", tag = "v0.2.0" }
switchyard-protocol = { git = "https://github.com/NVIDIA-NeMo/Switchyard.git", tag = "v0.2.0" }
tokio = { version = "1", features = ["macros", "rt"] }

2. Construct an algorithm. Target names are whatever your harness calls its models. This is the stage router from the benchmark; random, llm_task_classifier, and llm_classifier are built the same way.

from switchyard.libsy import LlmResponse, Step
from switchyard.libsy.algorithms import stage_router

algorithm = stage_router(
    "capable",
    "efficient",
    picker="efficient_first",
    confidence_threshold=0.5,
)

3. Drive it. run_stream takes an OpenAI-style request dict and yields steps. A CallModel step is a classifier or judge call — make it with your own client and hand the result back. Done carries the pick.

async def route(request: dict, clients: dict) -> tuple[str, dict]:
    async for step in algorithm.run_stream(request):
        match step:
            case Step.CallModel(call):
                target = call.models[0]
                try:
                    response = await clients[target].call({**call.request, "model": target})
                except Exception as error:
                    call.fail(error)
                else:
                    call.respond(LlmResponse.Agg(response))
            case Step.Done(outcome):
                return outcome.selected_model_ids[0], outcome.request

clients is your existing per-model client map. call.models lists fallbacks in order; outcome.request is the request to send, which may carry a rewrite the algorithm applied.

4. Make the answer call with the returned model and request, using your own HTTP client, retries, and credentials. If outcome.response is set, routing already produced the answer and you can return it directly.

Type reference: switchyard-libsy and switchyard-protocol. In Rust the loop is Algorithm::run_stream yielding Step::CallModel and Step::Done, with switchyard-llm-client's run available to drive it for you.

Path 3 — Run the Standalone Proxy

You finish with a server on localhost:4000 that any OpenAI or Anthropic client can call. Needs Rust with Cargo.

1. Install the server.

cargo install --locked switchyard-server

2. Write routes.toml. A stage router over the same model pair as the benchmark above: how to reach a provider, which models to use, how to choose between them. --config takes any path; this writes it to the current directory.

cat > routes.toml <<'TOML'
schema_version = 1

[llm_clients.openrouter]
format = "openai_chat"
base_url = "https://openrouter.ai/api/v1"
api_key_env = "OPENROUTER_API_KEY"

[targets.capable]
id = "anthropic/claude-opus-4.8"
llm_client = "openrouter"

[targets.efficient]
id = "z-ai/glm-5.2"
llm_client = "openrouter"

[routes.switchyard]
id = "switchyard"
type = "stage_router"
capable_target = "capable"
efficient_target = "efficient"
picker = "efficient_first"
confidence_threshold = 0.5
TOML

Every key is documented in the server configuration guide.

3. Start it. --dry-run loads the config, prints server OK: and the model IDs it exposes, then exits without starting the server.

export OPENROUTER_API_KEY="your-openrouter-key"  # pragma: allowlist secret
switchyard-server --config routes.toml --dry-run
switchyard-server --config routes.toml --host 127.0.0.1 --port 4000

4. Send a request. The route's id is the model name clients ask for.

curl http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"switchyard","messages":[{"role":"user","content":"hello"}]}'

The same route also answers on /v1/messages (Anthropic Messages) and /v1/responses (OpenAI Responses). /v1/stats reports which target served what, and /metrics exposes Prometheus counters for requests, errors, latency, tokens, and routing overhead.

5. Point a coding agent at it.

export ANTHROPIC_BASE_URL="http://localhost:4000"
export ANTHROPIC_MODEL="switchyard"
claude

Codex CLI and other OpenAI clients use the OpenAI variables instead:

export OPENAI_BASE_URL="http://localhost:4000/v1"

Routing Algorithms

Most use an LLM as a judge. All of them pick between an efficient model and a capable one; what differs is when the decision is made and how.

Algorithm	How it decides	Route type	Benchmark
Capability	The first request is judged by an LLM.	llm_classifier	71.2% at $79.32
Stage	Tool responses are judged by pattern matching or an LLM.	stage_router	72.7% at $68.19
Capability + Stage	Combines the two above.	composite	not yet benchmarked
Escalation	Starts efficient. Responses are judged by an LLM for issues, then escalated.	llm_classifier + mode = "escalation"	75.7% at $85.00
Advisor Gate	One model serves every turn; a stronger advisor approves its plans and "done" claims, or sends it back.	advisor	lifts a weak executor 43.8% → 54.7%
Sub-Agent-Aware	Delegated sub-agent traffic routes separately from the parent agent.	subagents on passthrough or stage_router	not yet benchmarked
Custom	The first request is judged by an LLM against criteria you define, routing among 2+ of your own models.	llm_classifier + target_selector policy	not yet benchmarked
Random	Each request is routed at random, uniform or weighted.	random	baseline mechanism

Benchmarks are Terminal-Bench 2.1 against a $98.06 Opus 4.8 baseline at 76.0%. A passthrough route registers one target under one model ID with no routing decision. See the Routing Overview for the common route shape and self-hosted targets.

Documentation

• Core Concepts: LLM clients, targets, routes, model IDs, and routing algorithms
• Routing Overview: choose and configure a routing algorithm
• TOML Schema: every configuration key
• Architecture: how the proxy and library components fit together
• switchyard-server: server configuration, routing algorithms, and metrics
• switchyard-libsy: embed routing algorithms in a Rust application
• switchyard-protocol: provider-neutral request, response, and streaming types
• switchyard-translation: request, response, and stream translation
• switchyard-nemo-relay-plugin: install Switchyard as a native NeMo Relay plugin

Benchmark Provenance

Configuration	Accuracy	Total cost	vs. Opus 4.8 baseline
Opus 4.8 baseline	76.0%	$98.06	—
Escalation	75.7%	$85.00	99.6% of accuracy, 13.3% cheaper
Stage	72.7%	$68.19	95.7% of accuracy, 30.5% cheaper
Capability	71.2%	$79.32	93.7% of accuracy, 19.1% cheaper
Kimi K2.6 alone	55.8%	$76.28
GLM 5.2 alone	52.4%	$16.47
DeepSeek V4 Pro alone	48.7%	$96.92
Ultra 3 alone	39.0%	$29.66

These are the v0.2.0 Terminal-Bench 2.1 results from Route AI Agent Workloads Across Models with NVIDIA NeMo Switchyard. Those runs used NVIDIA-internal inference endpoints, so absolute solve rates may shift on another serving stack; the routing parameters are the ones that ran.

The escalation deployment is checked in at benchmark/routing-profiles/tb21-escalation-opus-glm-deepseek.toml, with OpenRouter targets substituted so it is publicly runnable. To run the harness, see benchmark/README.md; for latency and routing overhead rather than task success, see Soak Testing.

Community

• Issues: GitHub Issues
• Code of Conduct: Code of Conduct

License

Apache 2.0 License. Copyright NVIDIA Corporation.