Every AI model inside Codex
Add one MCP server to Codex and it can call every model on the catalog: ask GPT, Gemini, Grok or DeepSeek for a second opinion with compare, search the web, transcribe audio, embed text, and make images, video and speech with the files handed back. One remote URL, one prepaid key, set up with four lines in config.toml.
Add the server to Codex
The server is https://api.relayswitch.io/mcp, Streamable HTTP, with your key sent as a Bearer header. Keys start sk-relay-.
~/.codex/config.toml
[mcp_servers.relayswitch]url = "https://api.relayswitch.io/mcp"bearer_token_env_var = "RELAYSWITCH_KEY"Codex sends Authorization: Bearer with the variable’s value. The CLI, the IDE extension and the desktop app share this file. Doc: learn.chatgpt.com/docs/extend/mcp
Or from the terminal
codex mcp add relayswitch --url https://api.relayswitch.io/mcp \ --bearer-token-env-var RELAYSWITCH_KEYCheck it from a terminal
curl https://api.relayswitch.io/mcp \ -H "Authorization: Bearer $RELAYSWITCH_KEY" \ -H "Content-Type: application/json" \ -H "Accept: application/json, text/event-stream" \ -d '{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}'What the client can call
The tools this page is about. The server lists 15 in all; the rest are on the MCP server page.
| Tool | What it does | Key |
|---|---|---|
compare | The same question to 2 to 5 models at once: every answer side by side with its cost and time. | Bearer |
chat | A conversation to any chat model, with a system prompt, images for the models that see them, and web search. | Bearer |
search | One web search: titles, URLs, snippets and dates, up to 10 results. | Bearer |
generate | Runs any image, video, speech, music or 3D model and returns the files, a finished image inline. | Bearer |
transcribe | Speech in an audio or video file at a URL to text, or translated into English. | Bearer |
list_models | Every chat, embedding, rerank and transcription model with its type, the tool that runs it and its price. | none |
What each call costs, read live
From the catalog at render. The developer’s list price is shown where the catalog has it.
| Model | Input / M | Output / M | List input / M | List output / M |
|---|---|---|---|---|
| Gemini 3.1 Pro Previewgemini-3.1-pro-preview | $1.20 | $7.20 | $2.00 | $12.00 |
| Grok 4.6grok-4.6 | $0.60 | $1.80 | $2.00 | $6.00 |
USD per million tokens. compare bills each model it asks at its own rate.
| Model | Setting | Price | Developer’s list |
|---|---|---|---|
| Nano Banana 2nano-banana-2 | resolution 1K | $0.0402 per image | $0.067 |
| resolution 2K | $0.056 per image | $0.101 | |
| resolution 4K | $0.072 per image | $0.151 | |
| Veo 3.1 Fastveo-3.1-fast | resolution 720p | $0.48 per generation | $0.80 |
| resolution 1080p | $0.576 per generation | $0.96 | |
| resolution 4k | $1.20 per generation | $2.40 |
Charged when the model accepts the task; a failed generation is refunded in full.
What to ask once it is connected
Plain requests; the client picks the tool.
- Use compare to ask gpt-5.2, gemini-3.1-pro-preview and grok-4.6 to review this diff, then tell me where they disagree.
- Search the web for the latest release notes of this library and summarize what changed.
- Generate a 16:9 header image for the README with nano-banana-2, quote it first.
- Transcribe this meeting recording (paste its https URL) and pull out the action items.
- List the embedding models with list_models and pick one for this repo’s docs search.
Codex MCP: questions
Setup, models and billing.
Every model as a tool in your AI
From $10 by card or in USDT, USDC, BTC, ETH, SOL. Prepaid, no subscription, no KYC.