Khimba/ Codex

Codex integration

Ask another AI from Codex.

Khimba is a hosted remote MCP that lets Codex prepare repository context locally, show it for approval and ask the model you choose for an independent review.

codex mcp add khimba --url https://mcp.khimba.ai/tools/model-opinions
codex mcp login khimba

Try a focused request

“What does DeepSeek think of my implementation plan?” or “Ask GLM to review my uncommitted changes for correctness, concurrency risks and missing tests.” Codex resolves the model and focus, then prepares the relevant local evidence.

Approve before anything leaves

The preparation shows the exact model, question, paths, file count, token estimate and maximum dollar charge. Nothing is uploaded and no credit is reserved before explicit confirmation. If the requested model is unavailable, Khimba reports that rather than substituting another.

What returns

The completed opinion contains findings, assumptions, recommendations, token usage and final measured charge. Treat it as another reviewer’s input, not proof. Reproduce important claims with tests and source evidence.

Guests begin with USD 0.10 promotional credit and need no account, card, provider API key or local Khimba server. Pricing is measured official-provider cost plus a 30% platform markup.