deixic
Log inRequest a demo

Platform

Change the model. Keep the work.

Dex, thinking it through

Use OpenAI, Anthropic, Google, Bedrock, Vertex, xAI, DeepSeek, Kimi, Qwen, GLM, or anything on OpenRouter, and keep your context, skills, connections, tests, and the approval policy on every fix.

Request early access

  1. Budget checkWithin the finance team budget
  2. AnthropicRate limited, moving to the next provider
  3. OpenAICompleted
  4. Reviewer modelRead-only check passed
  5. SpendRecorded to finance-ops
One request during a provider rate limit.

  • Bring your own keys

    Model usage runs on your provider accounts by default.

  • Fast model, second opinion

    Pair a quick, inexpensive model with a read-only reviewer model, in profiles from low to ultra.

  • Provider outages do not stop the work

    When a provider errors or rate-limits, the gateway moves the call to the next provider in line.

  • Local models too

    Deixic Code runs models on your own machine through Ollama, LM Studio, or llama.cpp.

How a request reaches a model

How it works in detail

Shared by every surface.

Use it with Dex on every surface, alongside company context that existing agents read over MCP.