GLM-52 897
GPT-56SC 873
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861
GLM-5 856
MUSE-SPK 841
QWEN-38X 824 -2.3%
GPT-6A 820
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
GLM-52 897
GPT-56SC 873
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861
GLM-5 856
MUSE-SPK 841
QWEN-38X 824 -2.3%
GPT-6A 820
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
← Back to feed

OpenAI's First Hardware Is a $230 Keyboard That Shows You What Your Agents Are Doing

OpenAI has shipped Codex Micro, a $230 limited-edition desktop keypad designed to let developers monitor and control multiple parallel Codex agents from a physical interface. The device was built in collaboration with Work Louder, a Canadian-Italian boutique keyboard manufacturer known for compact programmable peripherals.

It is OpenAI’s first branded consumer hardware.

What It Does

Codex Micro is not a general keyboard. It is a dedicated agent status panel. A row of translucent keycaps at the top of the device cycle through distinct colors — each representing the live state of a running Codex thread. A green key means the agent is executing. Amber means it is waiting on input. Red means it has stopped or failed. The physical display eliminates the need to tab between browser windows to check on queued agents.

The rest of the keypad provides programmable shortcut keys for common agent operations: starting new threads, pausing runs, injecting context, and navigating the Codex CLI interface.

The device is targeted explicitly at what the product team is calling “agentmaxxers” — developers who run four, eight, or more Codex threads concurrently and need a low-latency way to triage which threads need attention without context-switching to a screen.

Why Hardware, and Why Now

The Codex Micro launch is a signal about where the coding agent workflow has landed. When a single developer is managing a handful of subagents simultaneously — each one running autonomously against a different task, branch, or codebase — the management overhead becomes its own problem. Browser tabs and CLI logs are built for sequential attention. Physical state indicators are not.

OpenAI has watched Codex usage scale past its original assumptions. The Codex CLI, released in April 2026, was designed for single-agent use. Claude Code Dynamic Workflows and multi-agent harnesses pushed the paradigm toward dozens of parallel threads. The Codex Micro keypad is OpenAI’s acknowledgment that the peripheral layer matters at that scale.

Work Louder brings existing credibility in the mechanical keyboard community. The company’s prior products have focused on programmable macropad form factors, and Codex Micro extends that aesthetic into an AI-native interface with RGB state signaling.

Pricing and Availability

The Codex Micro is priced at $230. It is listed as a limited edition, with no public inventory count disclosed. Orders opened alongside the announcement. Shipping timeline has not been confirmed for all regions.

No API integration beyond the existing Codex CLI is required. The device connects via USB and uses configurable key mappings to trigger existing Codex agent commands.

What It Tells You About the Market

The $230 price point and limited edition positioning keep this in the enthusiast tier. It is not a mass-market product. But the categories of developer that would buy it — engineers running Codex at scale for production code generation — are also the developers generating the bulk of OpenAI’s API revenue.

The hardware serves a dual function: it gives power users a tangible workflow upgrade and makes the Codex brand physically present in the workspace. Both outcomes matter to OpenAI as coding agent competition from Anthropic (Claude Code), GitHub (Copilot CLI), and Cursor continues to tighten.

No other frontier AI lab has shipped a hardware product of any kind. OpenAI is first.