Conan is a native macOS application that functions as a Claude Code cockpit, providing a live heads-up display (HUD) for your AI coding sessions. It is purpose-built for developers, AI engineers, and power users who run Claude Code on Apple silicon Macs. The core value proposition is real-time observability: instead of squinting at a scrolling terminal, users see every prompt, tool call, skill activation, and token burn as it happens. This transforms the chaotic terminal output into a structured, streaming timeline and context monitor, giving users unprecedented insight into their AI agent's behavior. Conan turns the opaque coding session into a transparent, controllable experience.
The main pain point Conan solves is the lack of visibility into Claude Code's inner workings. When running an AI agent in a traditional terminal, actions stream by rapidly with no way to track what the agent did, how many tokens it used, or which skills it invoked. This makes debugging costly, context management guesswork, and collaboration difficult. Conan addresses this by surfacing every event in a structured HUD: users can see exactly which tools were called, how the context window filled, and when the agent paused. This clarity reduces wasted tokens, prevents context overflows, and enables users to understand their agent’s decision-making process.
The first major feature is the Streaming Timeline, which shows every command, edit, and tool call as it happens. Each event appears on a living timeline with timestamps, so users can scroll back to review past actions. This eliminates the need to grep through terminal history or replay sessions. The timeline also displays skill activations, subagent events, and user prompts. By presenting the entire session as a chronological feed, Conan makes it easy to pinpoint exactly where an agent went wrong or why it took a particular action.
The second feature group combines Context Monitor and Pulse. The Context Monitor breaks down the context window into system, tools, memory, skills, and messages, showing real-time token consumption. Users see how many tokens each category uses and track cumulative burn rates. The Pulse is a live throughput meter that spikes when Claude is actively working and calms when it waits. Together, these features let users anticipate when they need to start a fresh session or trim context. They also reveal the cost of each operation in both tokens and time.
The third feature group is Skills & MCP Observer and Claude Radio. The Skills & MCP panel surfaces every skill and MCP server currently in play, showing their last fired time and file path. This includes built-in skills like frontend-design, automate-browser, scan-trends, and handoff. Users can verify which skills are active and how they were triggered. Claude Radio provides a built-in music player with curated stations (Claude FM, System Of A Down orchestral, etc.) and support for any YouTube URL. The radio hums in the background, helping users maintain focus during long coding sessions.
admin
Conan works by wrapping Claude Code in its own terminal and capturing all output via the API. It does not proxy or modify Claude Code – it only observes. The app runs as a native macOS process, displaying a persistent HUD overlay that can be toggled or resized. All data stays local, and Conan does not send code or data to external servers. It is designed to have minimal performance impact, as it only reads events already generated by Claude Code. The HUD updates in real time, with no polling or delays.
Concrete use cases include debugging a custom skill: when automate-browser fails, users see the exact prompt and tool call that led to the error. For token management, developers watch the context monitor to decide when to perform a handoff using the handoff skill, which writes a structured HANDOFF.md for a fresh session. The scan-trends skill lets researchers gather data from Reddit, X, Twitter, YouTube, Hacker News, Polymarket, and the web over configurable time windows, with full visibility into each step. Frontend designers can run frontend-design skill and monitor its prompt history to refine results.
Conan targets developers and AI practitioners using Claude Code on macOS 13+ with Apple silicon. It is free to download, with a one-time Premium license of $29 for lifetime access to version 1.x. The app is not affiliated with Anthropic but is built by independent developers. It supports both free and paid Claude Code accounts, and does not require any special configuration beyond having Claude Code installed. With its real-time HUD, Conan empowers users to command their AI campaigns with full visibility, ensuring they never miss a move.
Conan is designed for developers and AI practitioners who use Claude Code on macOS 13+ with Apple silicon. This tool is ideal for software engineers, frontend designers, data researchers, and power users seeking real-time visibility into AI coding sessions. It serves teams leveraging Claude Code for automated workflows who need to monitor token usage, tool calls, and skill execution. Users who rely on custom skills and MCP servers will benefit from Conan's observability features. The app also appeals to solo developers who want a native, non-intrusive way to understand and optimize their AI agent interactions without performance overhead.
Updated 2026-06-15