How It Works
Cross-Code Organizer (CCO) is a local control panel designed to manage configurations across multiple AI coding harnesses. It consolidates visibility and control over tools such as Claude Code, Codex CLI, OpenCode, and DeepSeek Harness (DSH). The tool operates on a principle of normalization, where each harness retains its own rules, paths, and capabilities, but CCO provides a unified interface to inspect, edit, and clean these configurations without requiring users to navigate hidden directories or understand complex precedence rules.
The core functionality of CCO revolves around a three-step process: Scan, Find, and Fix. During the Scan phase, CCO identifies and lists all active components—such as memories, skills, agents, hooks, sessions, and MCP servers—across the supported harnesses. The Find phase reveals what each harness actually loads, including inherited settings, duplicates, and potential security risks. Finally, the Fix phase allows users to edit configurations directly within the dashboard, move misplaced items, delete duplicates, or address security findings by navigating directly to the relevant MCP server entry.
CCO achieves this through a modular architecture where each harness is handled by a dedicated adapter. These adapters normalize the behavior of different tools, ensuring that users do not need to understand the underlying precedence or mutation rules of each harness. This cross-harness visibility is further enhanced by features like the Effective Context Map, which shows how context is accumulated across directories and files, and the hygiene score, which evaluates the overall health of the configuration setup.
A key innovation in CCO is its Activation Scanner research preview, which introduces a novel method for detecting potential poisoning in MCP servers, skills, plugins, hooks, and tool descriptions. Unlike traditional scanners that rely on text, rules, or classifiers, this feature uses a frozen local sensor model to read hidden activations and train a probe before executing untrusted capabilities.
CCO also addresses common pain points in AI coding environments, such as wasted context tokens and misconfigured MCP servers. The tool provides a Context Budget feature to track token usage across memories, skills, and other artifacts, helping users reclaim wasted resources. Additionally, it allows users to disable MCP servers on a per-project basis, reducing the risk of unintended data exposure or malicious activity.
CCO only reads selected configuration files on the user’s machine, such as those in ~/.claude/, ~/.codex/, and project-level directories. Telemetry is entirely opt-in and anonymous, with no upload of paths, names, prompts, file contents, sessions, or credentials. Detailed event logs remain local, and any submitted data includes only a rotating anonymous ID, UTC month, app version, and harness ID.
CCO can be installed and run directly via npm using the command npx @mcpware/cross-code-organizer. On first launch, it auto-installs a /cco skill for Claude Code users, while Codex users can switch harnesses from the sidebar without additional setup. The dashboard interface is intuitive, with a project list, MCP servers displayed with security badges, and a detail inspector for deep dives into configurations. Users can click on security findings to navigate directly to the affected MCP server, making it easier to address issues without switching tools.
The project’s creator identified over 140 invisible configuration files controlling these tools and sought to eliminate the need for manual inspection. Since its release, CCO has gained significant traction, achieving over 100 stars in its first five days, reflecting its utility in streamlining AI coding workflows and enhancing security.
Use Cases
Cross-Code Organizer is primarily used to manage and secure configurations across multiple AI coding harnesses, making it particularly valuable for developers, security professionals, and teams working with AI-assisted coding tools. One of its key use cases is configuration hygiene, where users can identify and eliminate duplicate memories, stale instructions, or misplaced configurations that consume valuable context tokens. For example, a developer working with Claude Code might discover that multiple AGENTS instructions or inherited project files are bloating their context window, reducing the effective capacity for their actual coding tasks.
Another critical use case is security scanning and mitigation. CCO’s Activation Scanner research preview allows users to detect potential poisoning in MCP servers, plugins, or tool descriptions before they are executed. This is particularly important in environments where untrusted or third-party tools are integrated, as compromised servers can embed hidden instructions that exfiltrate sensitive data or alter behavior. By providing a direct path to inspect and disable suspicious MCP servers, CCO helps mitigate these risks without requiring users to switch between multiple tools or interfaces.
CCO also excels in cross-harness migration and compatibility. Developers often work with multiple AI coding tools, each with its own configuration format and precedence rules. CCO simplifies this by offering a unified dashboard to manage Claude Code, Codex CLI, OpenCode, and DeepSeek Harness configurations. For instance, a user might need to migrate a skill from Claude Code to Codex CLI or adjust settings across different projects.
The tool’s Harness Doctor feature provides a scope-level audit of the entire configuration setup, offering an Effective Context Map that visualizes how context is inherited across directories and files. This is useful for diagnosing why certain behaviors or errors occur, as it reveals direct, ancestor, and global sources of configurations. The explainable hygiene score further breaks down deductions into visible findings, such as duplicate identities, missing metadata, or large context-bearing files, helping users prioritize cleanup efforts. This diagnostic capability is particularly valuable for teams maintaining complex AI coding environments.
This is useful for maintaining performance and storage efficiency in long-running AI coding sessions.
For developers working with MCP servers, CCO offers granular control over their enablement and disablement on a per-project basis. This is important for isolating potentially risky or untrusted servers while still allowing access to trusted ones. The tool’s ability to navigate directly from a security finding to the relevant MCP server entry streamlines the process of investigating and remediating issues, reducing the time and effort required to maintain a secure coding environment.
Key Features & Capabilities
Cross-Code Organizer distinguishes itself from other configuration management tools through a combination of unique features and capabilities. One of its standout offerings is the Effective Context Map, which provides a detailed breakdown of how context is accumulated across different directories and files for each harness. This feature is particularly useful for understanding why certain configurations are loaded or why context windows are filled inefficiently. Unlike standalone scanners or desktop apps, CCO ties this visibility directly to actionable fixes, allowing users to edit configurations in place or move items to their correct locations.
Another key feature is the Activation Scanner research preview, which introduces a novel approach to detecting poisoning in MCP servers and other components. Traditional scanners rely on text analysis, rules, or classifiers, but this preview uses a frozen local sensor model to read hidden activations and train a probe before executing untrusted capabilities. This method can identify subtle manipulations, such as embedded instructions in tool descriptions, that might otherwise evade detection. The scanner’s findings are directly actionable, with users able to navigate to the affected server or component and take corrective measures.
CCO’s Harness Doctor provides a comprehensive audit of the configuration setup, offering an explainable hygiene score that evaluates the overall health of the environment. This score is derived from visible findings, such as duplicate identities, stale sessions, or large context-bearing files, and serves as a diagnostic heuristic rather than a security certification. Users can initiate safe auto-repair for byte-identical artifacts, with CCO archiving the extra copy and providing undo functionality.
This is critical for optimizing the use of limited context windows, as duplicates or inherited settings can consume tokens unnecessarily. By identifying and removing redundant configurations, users can reclaim wasted resources and improve the efficiency of their AI coding sessions. Additionally, CCO allows users to disable MCP servers on a per-project basis, further reducing the risk of unintended data exposure.
Cross-harness support is another defining feature of CCO, with dedicated adapters for Claude Code, Codex CLI, OpenCode, and DeepSeek Harness. Each harness retains its own rules and capabilities, but CCO normalizes their behavior to provide a unified interface. For example, Claude Code’s memories and skills are displayed alongside Codex CLI’s AGENTS instructions and profiles, enabling users to manage multiple tools from a single dashboard. This cross-harness visibility extends to features like the All Memories view, which includes body search and full Markdown editing capabilities.
This is particularly useful for maintaining large or complex configuration setups, where manual adjustments could be error-prone or time-consuming.
Security is a central focus of CCO, with features designed to detect and mitigate risks associated with MCP servers, plugins, and tool descriptions. The tool’s ability to scan for poisoned servers and navigate directly to affected entries streamlines the remediation process. This not only improves performance but also minimizes the attack surface by reducing the volume of data that needs to be managed.
The tool is designed for zero-install usage, with support for running directly via npx @mcpware/cross-code-organizer. This eliminates the need for complex setup procedures or dependencies, making it accessible to users who prefer lightweight, on-demand tools. The dashboard interface is intuitive, with a project list, security badges for MCP servers, and a detail inspector for deep dives into configurations. This design ensures that users can quickly identify and address issues without needing to switch between multiple tools or interfaces.
CCO’s development has been driven by a focus on addressing real-world pain points in AI coding environments. The project’s creator identified over 140 invisible configuration files controlling these tools and sought to eliminate the need for manual inspection or editing. By providing a unified dashboard for configuration management, security scanning, and cleanup, CCO aims to improve the efficiency, security, and usability of AI-assisted coding workflows.