sapiom router

Reduce your agent costs in minutes.

Point any OpenAI-compatible agent at one endpoint. Sapiom selects the most efficient allowed model for each request while keeping policy and cost visible.

Start building free Read Router docs

OpenAI-compatibleone API keyper-request metering

sapiom router chat completions

curl https://llm.services.sapiom.ai/v2/openai/v1/chat/completions \
  -H "Authorization: Bearer $SAPIOM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      {"role": "user", "content": "Say hello from Sapiom."}
    ]
  }'

model routing one endpoint

task shapeinput sizecontext sizelatencyqualitycost

company policyreliabilityavailabilityaudit trailallowed modelsruntime health

efficient modelmid-tier modelfrontier modelretry routefallback routepolicy hold

Keep the interface. Change the economics.

Send the same chat-completions request to Sapiom. Router evaluates the task at execution time and selects the best allowed path under your cost and policy constraints.

Change one endpoint.

Call the Sapiom chat-completions URL with your API key. The request stays compatible with the OpenAI message format your agent already uses.

Read the model guide

Router endpointPOST

protocolOpenAI compatible

path/chat/completions

authorizationBearer $SAPIOM_API_KEY

content typeapplication/json

Keep the request shape.

Your agent continues sending a standard messages array. No capability discovery, custom tool name, or rewritten application flow is required.

Read the API docs

{
  "messages": [
    {
      "role": "user",
      "content": "Say hello from Sapiom."
    }
  ]
}

Route at execution time.

For each request, Router weighs task requirements against cost, quality, latency, reliability, availability, and company policy before choosing an allowed model.

See monitoring and controls

Runtime decisionper request

simple taskefficient model

mid-weight taskmid-tier model

hard taskfrontier model

every callmetered + recorded

From request to routed response.

Point to Sapiom

Send your existing OpenAI-compatible request to the Router endpoint with one Sapiom API key.

Route each request

Sapiom chooses the most efficient allowed model for the task at the moment it runs.

See every call

Cost, route, and execution metadata stay metered and recorded for the agent behind the request.

Route your first request.

Use one OpenAI-compatible endpoint and keep the request body your agent already sends.

Start building free Read the docs