Connect your agent
Use a compatible client’s base URL setting or the planned local daemon. Keep the protocol your tool expects, with a scoped key for each workspace.
Keep your preferred coding workflow. Enable permitted data capture for a later sale, or choose an eligible program that lowers inference costs as you work.
Gateway-assisted example. Direct BYOK can use supported capture hooks without rerouting inference. Live routing is not yet available.
Your workflow enters one policy-aware route. OptinRouter selects an eligible model and keeps the response connected to its request.
OptinRouter is designed to sit between your agent and its model provider, with the controls and context you need in one place.
Use a compatible client’s base URL setting or the planned local daemon. Keep the protocol your tool expects, with a scoped key for each workspace.
Choose model preferences, budgets, and permitted routes. Organization rules take priority, including the option to keep every repository Private.
The gateway links permitted requests and replies. Supported hooks add tool results, tests and outcome evidence, with capture and sale permissions kept separate.
Start with a hosted connection, or add a local layer when you need protection and session context on your own machine.
Explore the integration designPoint a compatible client at the gateway with a scoped key. A simple path for remote workspaces, tools, and CI.
Add local secret filtering, repository policies, and device-signed work context before a request or submission leaves your machine.
The planned platform captures and organizes work only with permission. Keep your existing inference arrangement and prepare data for a later sale, with optional brokerage, or accept a qualifying data-for-discount agreement before use. Approved benefits may include inference discounts, nonwithdrawable credits or direct revenue.
Understand the consent flowChoose permitted repositories and fields. Keep private work outside capture.
Sell prepared data later, or review a qualifying agreement before use.
Specify buyers, uses and benefits. Automatic delivery follows only accepted rules.
No. Private is the planned default. Routing a request to a model provider is separate from submitting a session to the contribution program. Your provider still receives the input required to answer your request under its own terms.
Yes, this is a core design requirement. Organization policy can lock the workspace to Private and restrict which repositories are ever eligible for contributions. These controls are part of the planned gateway release.
Compatibility depends on whether a tool allows a custom endpoint and on its protocol features. The initial focus is OpenAI-format and Anthropic-format clients. Tested tool versions and known limitations will be published before production support is claimed.
This website and the interactive workspace demo are available to explore. The inference gateway, local daemon, account billing, and contribution program are in development. Request early access to discuss your workflow.
Help shape a model gateway built around the way developers actually work.