Claude Cowork
How to run Claude Cowork with third-party inference providers
A walkthrough using Opper as the gateway: configure Cowork's Developer-mode dialog, add models across AWS Bedrock, GCP Vertex, and Azure, and switch providers mid-session.
Agent CLI
Run Claude Code, Codex,
and more on any model
A CLI router for AI agents, including a hosted Claude Code router for any model. Switch models mid-task, fall back across providers, all on one OAuth login.
Why a gateway
Residency, fallbacks, cost
EU data residency
Every model in the catalog is tagged with where it's deployed. Configure your agent to route only to EU-hosted models (Claude on GCP Vertex EU, Sonnet on AWS Bedrock EU, open-weight on Evroc) when inference has to stay in Europe.
Browse the model catalog →Automatic fallbacks
The same model (Claude Sonnet, say) is wired up across AWS, Azure, and GCP. If one rate-limits or goes down mid-session, the gateway routes to the next upstream without interrupting the agent.
How fallbacks work →Cheaper for routine work
Not every prompt needs Opus. Run OpenCode or OpenClaw on a smaller or open-weight model for routine work (up to 98% cheaper for the same task) without leaving the CLI.
See the cost breakdown →Security
Built for sensitive work
Your data stays yours
Opper never trains on customer data: code, emails, documents, or anything else your agent reads. Optional zero data retention drops prompts and completions at request time.
One scoped key, instant revoke
Log in once with OAuth. No OpenAI, Anthropic, or Google keys on your laptop. Every agent shares a single Opper-scoped key you can revoke from the dashboard at any time.
Encrypted, isolated, EU-hosted
TLS in transit, AES-256 at rest, isolated data per organization. Platform deployed in AWS Stockholm with daily encrypted backups.
Install
Install the Opper Agent CLI
One npm install, one OAuth login. Then launch Claude Code, Codex, OpenCode, OpenClaw, Hermes, or Pi.
terminal
# Install (Node 20.12+)
# Sign in (OAuth device flow)
# Launch an agent
# Or wire your editor
FAQ
Claude Code & Opper Agent CLI FAQ
How do I change the model or use a different LLM in Claude Code?
+
Install the Opper CLI, run
opper login, then launch Claude Code with any model:$ opper launch claude --model openai/gpt-5.5 $ opper launch claude --model anthropic/claude-opus-4-7Switch models mid-session by relaunching with a different
--modelflag. No config changes needed.How does this compare to claude-code-router or OpenRouter?
+
Claude Code Router is an open-source proxy you run and maintain locally. OpenRouter is a hosted inference gateway. The Opper Agent CLI is a hosted CLI that wraps both: install once, log in once, then 'opper launch <agent>' routes Claude Code (and Codex, OpenCode, Hermes, Pi, OpenClaw) through Opper's gateway, with EU residency, automatic fallbacks, and pay-as-you-go billing. No local proxy to maintain.
How is this different from running each agent directly?
+
Most agents need their own provider keys, billing accounts, and rate-limit handling. With Opper you log in once and every agent shares the same key, the same 300+ model catalog, and the same observability, so switching from Claude Code to Codex to Hermes doesn't mean re-configuring your environment.
Do I need separate API keys for OpenAI, Anthropic, etc.?
+
No. Every supported agent runs through Opper's gateway, which routes to 300+ models with one Opper API key.
Is Claude Code free? How much does it cost via Opper?
+
The Opper Agent CLI is free. Inference is pay-as-you-go: provider cost + 3% gateway fee, billed to your Opper account. Claude Code via Opper costs the same as Anthropic's direct rate plus a small gateway margin. See full pricing.