Conan

Product Introduction

1. Definition: Conan is a native macOS application (category: Developer Tools, AI Interface) that provides a real-time, graphical Heads-Up Display (HUD) for sessions running in Anthropic's Claude Code. It functions as a comprehensive observability layer, wrapping the command-line agent to visualize every interaction.
2. Core Value Proposition: Conan exists to solve the opacity of complex AI agent workflows. It delivers complete session transparency, allowing users to monitor every prompt, tool call, skill execution, and token consumption from Claude Code in a live, unified interface, turning a terminal-based process into a manageable campaign.

Main Features

1. Live Timeline & Event Streaming: The core feature is a real-time, chronological timeline that logs every event from the Claude Code session. This includes all PROMPT inputs, TOOL calls (like Bash or file operations), SKILL invocations (e.g., /handoff, scan-trends), and system events (STOP, NOTIF). The timeline provides a complete audit trail, showing exactly what the AI agent is doing as it happens, eliminating the need to scroll back through a terminal buffer.
2. Context Window Monitoring: A dedicated dashboard visualizes the active context window's composition and fill level in real-time. Users see token consumption broken down by category: System instructions, Tools definitions, Memory, Skills, and Messages. This includes live percentage readouts and a total token count (e.g., "621k / 1.0M"), enabling precise tracking of how close the session is to its context limit, a critical metric for managing long-running or complex tasks.
3. Session Pulse & Throughput Visualization: A dynamic "pulse" graph displays the throughput of the session, showing spikes when Claude is actively processing or executing tools and lulls when it is awaiting user input or idle. This provides an immediate visual cue for the agent's workload and activity state, helping users understand pacing and identify bottlenecks.
4. Skills & MCP Server Visibility: Conan surfaces the complete inventory of active skills (custom slash commands like /handoff) and Model Context Protocol (MCP) servers available to the agent. It lists each skill's name, a brief description, its last execution timestamp, and its file path. This demystifies the agent's capabilities, showing exactly "what tools it wields" at any moment.
5. Built-in Claude Radio: An integrated ambient audio player streams curated lo-fi and instrumental music playlists to enhance focus. Curated stations include genre-based orchestral arrangements (e.g., "Dr. Dre — Symphonic Orchestra Instrumentals") and focus-oriented mixes ("Claude FM"). The player accepts any YouTube URL, allowing users to customize the session's soundtrack with a session-specific link.

Problems Solved

1. Pain Point: The fundamental problem is the lack of transparency and control when using advanced AI coding agents via a command line. Key issues include difficulty tracking a long sequence of tool calls, unexpected context window exhaustion, inability to easily audit AI actions, and general session complexity leading to reduced productivity.
2. Target Audience: Power users of Claude Code, including AI developers, full-stack engineers, technical project managers, and researchers who engage in complex, multi-step AI-assisted coding or problem-solving sessions that generate extensive logs and context.
3. Use Cases: Essential for debugging intricate AI workflows where understanding the exact sequence of operations is critical; auditing AI-generated code changes for compliance or security; optimizing prompts and skill usage by analyzing token consumption patterns; maintaining focus during long generative sessions; and creating a reproducible record of AI-assisted work for documentation or handoff.

Unique Advantages

1. Differentiation: Unlike using Claude Code directly in a terminal (which requires manual log parsing and provides no context visualization) or generic code editors, Conan provides a purpose-built, native visualization layer specifically for the AI agent's internal operations. It is a passive observation tool that does not interfere with or alter the underlying Claude Code process.
2. Key Innovation: The key innovation is the creation of a native macOS HUD that passively monitors and visualizes the entire lifecycle of a Claude Code session in real-time. It treats the AI session as a complex system to be observed, innovating on the concept of an "agent cockpit" by providing deep telemetry on context, skills, and throughput without requiring integration into the agent's own toolchain.

Frequently Asked Questions (FAQ)

1. What are the system requirements for Conan? Conan requires macOS 13 (Ventura) or later running on Apple Silicon (M1/M2/M3/M4 series) Macs. It is not available for Intel-based Macs.
2. Is my code or data sent to Conan's servers? Does it slow down Claude Code? No. Conan is a native, on-device application that only observes the local Claude Code session. It does not send your code, prompts, or data to any external server. It operates as a separate, passive monitoring process and does not add overhead or slow down the core Claude Code agent's execution.
3. Do I need a separate Claude Code subscription to use Conan? Yes. Conan is a visual frontend and monitoring tool that requires you to run Claude Code separately. It is a one-time purchase ($29) for the app itself, while your Claude Code usage and subscription remain with Anthropic.
4. How does Conan differ from just watching the output in the Claude Code terminal? The terminal shows only the latest output and requires scrolling to review history. Conan provides a searchable, structured, and categorized timeline of all events, live context window analytics, skill inventories, and throughput metrics—all in a single, persistent dashboard that persists after the terminal session ends.
5. Is Conan affiliated with Anthropic, the company behind Claude? No. Conan is an independent, third-party application developed by an unrelated team. It is a commercial product that enhances the experience of using Anthropic's Claude Code but is not officially produced, endorsed, or affiliated with Anthropic.