Skip to main content
Concentrate AI works with the AI coding tools and desktop apps you already use. Because Concentrate is a drop-in, OpenAI-compatible gateway, most integrations are just two values (the base URL and your API key), which unlocks 150+ models from 20+ authors across 20+ providers without changing how you work.

Pick your tool

Match the tool you’re using to the guide that walks through the setup.

Claude Code

Anthropic’s terminal coding agent. 150+ models, no $200 subscription.

Claude Desktop

Route Claude Desktop through Concentrate via the third-party gateway.

Codex (Beta)

OpenAI’s Codex CLI, powered by any model on Concentrate.

Cursor (Beta)

Use Concentrate models in Cursor via OpenAI-compatible settings.

Agent skills

Teach your coding agent the Concentrate API with the open skills library.
More integrations are on the way. If your tool isn’t listed, it can almost certainly connect through the same OpenAI-compatible pattern: point the client at https://api.concentrate.ai/v1, use an API key that starts with sk-cn-v1-, and pick a model in provider/model-id format. See the Quickstart for a from-scratch setup.

What every integration has in common

Regardless of which tool you connect, the setup comes down to the same values.
Anthropic-native clients append /v1 to the base URL themselves, so leave it off. Setting ANTHROPIC_BASE_URL to the /v1 value produces /v1/v1/... paths and 404s. They also send the x-api-key header rather than Authorization: Bearer; Concentrate accepts both, prioritizing Bearer when both are present.

What you get

  • One key for every model. Access 150+ models from 20+ authors across 20+ providers through a single API key, with no per-provider accounts to manage. See the models for the current catalog.
  • Built-in observability. Usage, spend, latency, providers, models, and cache hits across your whole org, with no third-party integration.
  • Team-scale spend management. Budgets at the organization, team, developer, and key level.
  • Strategy-driven auto routing. Use model: "auto" with explicit optimization for cost, latency, or performance. See Auto Routing.

Feedback

If a tool didn’t connect cleanly, email support@concentrate.ai. We use integration friction reports to prioritize compatibility fixes.
Last modified on September 1, 2026