AI AGENTS

Connect compatible agents to RouterLab.

Use RouterLab as a model endpoint for agents and frameworks that let you configure a compatible provider or a custom API URL.

Compatible endpoint
Visible usage
Centralized credits
Connection

Two ways to connect your tools

Direct connection

Direct connection

No wrapper

For agents and frameworks that let you define a compatible provider or custom base URL.

Agent Zero
OpenClaw
Hermes
RouterLab
ScioNos Wrapper

Via ScioNos Wrapper

For clients that need a local adaptation layer before RouterLab.

Claude Code
Codex CLI
Claude Desktop
Claude Cowork
ScioNos Wrapper
RouterLab
Discover ScioNos Wrapper
Why RouterLab for agents?

Agents create loops. RouterLab makes their usage visible.

Request loops

An agent may make several calls to complete a single task.

Different models

A workflow can use different routes or model families.

Less visible cost

Without a common layer, consumption can be spread across several providers.

Centralized view

RouterLab brings routes, usage, and credits into one shared view.

Integration examples

Configure a compatible agent, one tool at a time.

These three tools are direct examples. Use their official setup flow, then enter the RouterLab endpoint in the fields they expose.

Agent Zero

Agent framework

Configure an OpenAI-compatible provider in Agent Zero. Model roles may require separate settings.

Installation
curl -fsSL https://bash.agent-zero.ai | bash
Configuration principle
ProviderOpenAI Compatible
API URLhttps://api.routerlab.ch/v1
API key<RouterLab API key>
Model<RouterLab model ID>

The exact configuration can depend on the model role used in Agent Zero.

Official Agent Zero guide

OpenClaw

Configurable self-hosted agent / assistant

Declare RouterLab as a custom provider without locking the OpenClaw adapter type here.

Installation
curl -fsSL https://openclaw.ai/install.sh | bash
openclaw onboard
Configuration principle
ProviderCustom provider
baseUrlRouterLab API URL
apiKeyRouterLab API key
modelRouterLab model ID

The exact mapping depends on the OpenClaw setup; this card shows the principle, not a ready-to-copy file.

Official OpenClaw guide

Hermes

Autonomous agent environment

Use the Custom endpoint flow, then enter the RouterLab URL, key, and model.

Installation
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes model
Configuration principle
ProviderCustom endpoint
Base URLhttps://api.routerlab.ch/v1
API key<RouterLab API key>
Model<RouterLab model ID>

Some Hermes workflows require a model with at least 64K context. Check Hermes requirements before choosing your route.

Official Hermes guide
Common principle

Three values to connect

Most of these integrations rely on a RouterLab API URL, a RouterLab API key, and a model ID.

Mapping diagram
3 fields
1ROUTERLAB_BASE_URL=https://api.routerlab.ch/v1
2ROUTERLAB_API_KEY=<RouterLab API key>
3MODEL_ID=<routerlab-model-id>

The exact parameter names depend on the agent. This block is a mapping diagram, not a universal configuration.

Another integration path

Do you use Claude Code or Codex CLI?

These clients use a different integration path. ScioNos Wrapper prepares the transport to RouterLab locally.

ScioNos is not needed for Agent Zero, OpenClaw, and Hermes in the direct examples above.

Discover ScioNos Wrapper

Ready to connect your agent?

Start with RouterLab, choose a model route, and keep your usage visible.

Start with the protocol. Choose a model family or a client integration only after that.