Model Dock For Codex

Work in one place.
Get things done everywhere.

Keep working in Codex. ModelDock brings the models you choose—native GPT, managed providers, and local engines—together with the tools a task may need, without manually moving prompts, files, or results between apps.

Download ModelDock ↓How it works ↓

LOCAL BRIDGE Responses streaming, tool loops and long-session work remain part of the same task.

ModelDock dashboard showing a Codex route, configured capabilities, and live task observability — A local bridge keeps the connected task observable.

01 / Model choice

Keep your way of working.
Add the routes you want.

ModelDock does not replace Codex with another chat window. It lets the model picker carry additional routes alongside native GPT models, so a task can use the right capability or cost profile without leaving the environment you already use.

Model picker showing native and additional routes in Codex — MODEL PICKER Native GPT, managed, and local routes—one familiar place.

02 / Capabilities

Give a model the capability the task calls for.

A fast text model can work with image understanding. A local model can search, remember, speak, hear, generate, or make an artifact when those tools are configured. Capabilities are added as tools inside the task—not as another application to coordinate.

EXAMPLES Vision · web search · memory · speech · image generation · video production

ModelDock dashboard showing configured model routes, capabilities, and task telemetry — ONE TASK Routes, tools, context and trace information belong together.

03 / Local models

Bring a local engine in.
Keep the task intact.

Ollama, llama.cpp, vLLM and compatible local endpoints can be detected on loopback or added by URL. ModelDock can show whether a local model fits, suggest practical settings, and make its request, context and latency behavior visible around the task.

Ollamallama.cppvLLMcompatible endpoint

Install the bridge →

Abstract schematic of a local inference machine connected to model endpoints — Local route available on loopback.

04 / Hardware evidence

Know the machine
before you build or buy it.

Local models have a machine behind them: artifact, quant, runtime, context, topology, power and the work you actually run. That makes the benchmark and purchase guide a natural ModelDock extension—turning local configuration into practical performance evidence and operating economics.

Explore benchmarks →View evidence data →

01 / HARNESS

Declared local run

Machine, model artifact, runtime, quant and context.

02 / MEASURE

Practical behavior

Prefill, decode, power and task conditions stay separate.

03 / EVIDENCE

Comparable source

Visible provenance and conditions before a result enters a table.

04 / ECONOMICS

Operating choice

Workload, time and energy translate behavior into guidance.

Work in one place.Get things done everywhere.

Keep your way of working.Add the routes you want.