From IDE add-ons to a new class of coding teammate
AI coding agents are software systems that can read and write code, interact with tools, and take multi-step actions across a development workflow, moving beyond simple autocomplete to behave like semi-autonomous collaborators. After a first wave of IDE plug-ins that added chat panels and code suggestions, a second wave is pushing these agents into new hardware and management layers. Instead of living only inside a single editor, they are turning into services that follow developers across devices and projects. This shift changes what a developer workstation looks like, and how teams think about AI agent management. IDEs still matter, but they are no longer the only home for agentic AI tools. Smart glasses, multi-agent dashboards, and protocol-based integrations are testing what happens when AI coding agents become part of the wider computing environment.
Kai shows IDEs evolving, but also their limits
Embarcadero’s Kai extension for RAD Studio shows how traditional IDEs are adopting agentic AI tools without rebuilding their core. Kai adds chat, code completion, and a Model Context Protocol server on top of Delphi and C++ Builder, so AI coding agents can communicate with the IDE to generate code, resolve build errors, manage version control, and perform file operations. According to The Register, Kai is a subscription product costing USD 249 (approx. RM1,150) per developer per year, even though users must bring their own cloud or local large language models via API keys. That structure keeps the IDE central, while treating AI as a configurable add-on. It reflects a conservative path: extend long-lived developer workstations rather than replace them. Yet Kai’s “minimalist” feel also hints that IDE-bound extensions may struggle to keep up as agents expand into more complex, cross-environment roles.
Smart glasses as wearable terminals for coding agents
Monako Glass pushes smart glasses development into developer territory by merging a Linux-based system, waveguide display, camera, and AI coding-agent connections into a 48-gram frame. Instead of pitching a full laptop replacement, Monako positions the glasses as a wearable command layer for tools like Claude Code, Codex, Unreal Engine, and Blender. The idea is that a developer can glance at an agent’s progress, approve steps, send prompts, or review output without going back to a desk. MonoOS, Monako’s Linux-based platform, adds a Lua application layer and embedded Rive animation runtime, while a bone conduction microphone and Vision Engine gestures handle input in noisy environments. For now, many details remain unknown, including chip, memory, storage, and real-world battery life. The most credible role is as a lightweight terminal for AI coding agents, extending developer workstations into always-on eyewear.
Devin Desktop and the rise of AI agent management layers
Cognition’s Devin Desktop treats AI coding agents as a fleet to manage rather than a single helper per developer. Built as the next generation of Windsurf, it combines a full code editor with a dashboard for coordinating local and cloud agents across projects and environments. Spaces let teams group agents by project and share context across pull requests, files, and tasks, so agent sessions stop feeling like isolated chats and start to look like persistent collaborators. Theodor Marcu of Cognition says, “The question for engineering leaders is no longer whether to use AI — it is how to manage a growing fleet of agents working across their organisation simultaneously.” Support for the Agent Client Protocol means third-party and internal agents, including tools like Codex and Claude Agent, can run inside Devin Desktop, while a planned agent router will route tasks based on performance and cost constraints.
From solo assistants to enterprise-grade coding infrastructure
Taken together, Kai, Monako Glass, and Devin Desktop show AI coding agents moving from narrow autocomplete features to shared infrastructure that reshapes developer workflows. IDE extensions like Kai keep agentic AI close to familiar tools but highlight the limits of treating AI as a side panel. Hardware experiments such as Monako Glass ask what happens when AI follows developers into every room as a wearable terminal. Coordination layers like Devin Desktop recognise that teams will run many agents—from different providers, on different machines—and need consistent AI agent management to keep work reliable and auditable. As these trends converge, the definition of a developer workstation broadens to include smart glasses, desktops, and cloud environments stitched together by agent-aware protocols. The ecosystem is diversifying beyond the IDE, pointing toward AI coding agents as first-class citizens in enterprise software engineering.
