corbits

Corbits Router

A smart gateway for AI model requests

Route AI requests according to your organization's policy. Enforce by task, cost, or policy before a request reaches inference.

Read the docs
Corbits Router — provider routing and credential injection Playback starts muted while the video is visible.

Route requests across providers, inject keys at the edge, and see what every call cost.

Enable world class inference without any of the operational headaches.

“I simply cannot deal with a lot of providers all the time.”
Prospect, sales call

  1. Provider credentials live in Corbits and are injected and routed at the edge, per request. No one at your organization has to switch providers ever.

    agent: support-agent

    provider_key: none

    injected_at: edge

  2. Route by task, cost, or policy, and fail over when a provider degrades.

    ClaudeOpenAIGeminiLocal

  3. Who asked, which model answered, and what it cost, on every request.

    $0.42

    finance/close-agent · claude · one request

Every request at your fingertips.

See which agent asked, which model answered, and what it cost.

router / requests

Part of Corbits

  • support-agentclaude · $0.012routed
  • research-agentgemini · $0.004routed
  • finance-agentopenai · over budgetblocked
  • release-noteslocal model · $0.000routed

FAQs

Does it work with agents we didn't build?
Yes. Agents on LangChain, Mastra, Agno, or your own code can send model requests through the router. Identity, grants, and the signed record come with agents running on Corbits.
Where do the keys live?
Encrypted at rest in Corbits and injected at the edge per request.
What about local models?
Any local server with an OpenAI-compatible API can act as a provider.

A smart request is the best request.

Read the docs