Cross-Code Organizer CLI
A local dashboard for Claude Code, Codex CLI and OpenCode: what the agent loads, its token cost and an MCP server security scan
- 387 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The dashboard reads agent configuration, memory and sessions in ~/.claude and ~/.codex and also connects to every configured MCP server to fetch tool descriptions
- It moves, edits and deletes settings, skill and memory files and compresses sessions, so confirm actions and rely on the built-in undo
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/cross-code-organizerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The dashboard reads agent configuration, memory and sessions in ~/.claude and ~/.codex and also connects to every configured MCP server to fetch tool descriptions
- It moves, edits and deletes settings, skill and memory files and compresses sessions, so confirm actions and rely on the built-in undo
This is third-party code. Review the repository files before installing.
What it does
Cross-Code Organizer, also CCO and formerly Claude Code Organizer, starts with `npx @mcpware/cross-code-organizer` and opens a local dashboard for your coding agents' settings. You pick an agent on the left (Claude Code, Codex CLI, OpenCode or DeepSeek Harness), and the dashboard shows what it sees in each scope: memories, skills, agents, commands, hooks, rules, sessions and MCP servers. Each agent has its own adapter that respects its own precedence rules. The working loop is scan, find, fix. The Show Effective mode reveals what actually applies in a project and marks shadowed items and name conflicts. The context budget counts tokens for every item, splitting always-loaded from deferred and expanding `@import`. Fixes are made right in the dashboard: you edit Markdown skill and memory files, move items between scopes and delete duplicates. Every action can be undone. The security scanner connects to every MCP server, fetches the real tool descriptions and runs them through 60 patterns and 9 deobfuscation techniques. SHA256 hashes show whether a server changed its descriptions between scans. There are also MCP tools for the agent itself, per-project server disabling, backups and distillation of bloated Claude Code sessions that keeps the original.
Who it is for. For people who work in Claude Code or Codex CLI and have piled up dozens of MCP servers, skills and memories and want to see what actually loads into context and remove the excess.
Good fit when
- Claude Code's context runs out faster than expected and you need to see how many tokens memories, MCP schemas and CLAUDE.md take
- The same MCP server is installed in several scopes and you need to find duplicates and remove extras
- Before connecting someone else's MCP server you want to check its tool descriptions for hidden instructions
- You want to disable some global MCP servers in one project without deleting them everywhere
- A Claude Code session has swollen from screenshots and long tool outputs and needs compressing before you continue
Not a fit when
- You need headless checks in CI: the `cco scan --json` command is on the author's roadmap, and for now this is a graphical dashboard
- Your agent is Cursor, Windsurf or Aider: support is planned, and today there are Claude Code, Codex CLI, OpenCode and DeepSeek Harness
- You need a hard blocker for malicious servers: the scanner raises warnings, and the activation probe mode is still a research preview
- You work on Windows and need scheduled backups: the scheduler does not work on that system
Example request
Run npx @mcpware/cross-code-organizer and tell me the URL when the dashboard is ready.Limitations
It needs Node.js 20 or newer. On Linux, macOS and WSL the dashboard is fully supported, on Windows 11 partially: the dashboard opens, but the backup scheduler does not. Backup scheduling relies on systemd on Linux and WSL and on launchd on macOS.
Support differs by agent. For Claude Code and Codex CLI edits and moves are available, for OpenCode and DeepSeek Harness only inventory for now, and unsupported changes stay disabled until the author models them safely. Some items, such as configuration and plugins, are visible but locked against moving.
The security scanner stays a heuristic: the author calls the hygiene score a diagnostic and not a security certificate. The activation scanner is a research preview and gives no final blocking. Metrics are off by default, and when enabled CCO sends once a month only an anonymous ID, the month, the version and the agent name. Details are in PRIVACY.md. Running it needs only access to npm.
How to disable. Stop the npx process and close the dashboard tab. Delete the /cco skill, which CCO installs on first run for Claude Code, from ~/.claude/skills/. CCO's local data lives in ~/.cco/ and can be deleted along with the backups if you do not need them.
FAQ
How do I start Cross-Code Organizer?
Run `npx @mcpware/cross-code-organizer`, open the address it prints and choose an agent in the sidebar. You can ask Claude Code or Codex CLI to run the command and report the address. On first run a /cco skill for Claude Code is installed.
How is it different from Claude Code Organizer?
It is the same project after a rename. It used to work only with Claude Code and now supports Claude Code, Codex CLI, OpenCode and DeepSeek Harness through adapters, with Cursor, Windsurf and Aider planned.
How do I check MCP servers for tool description poisoning?
Click the security scan button. CCO connects to every server, fetches the real descriptions and runs them through 60 patterns and 9 deobfuscation techniques, while NEW, CHANGED and UNREACHABLE badges mark new, changed and unreachable servers. A finding takes you to the server's entry.
How do I find out how many tokens Claude Code uses before work begins?
Open the context budget: it counts tokens for each item, splits them into always-loaded and deferred and expands `@import`. You can toggle between a 200K and a 1M window.
Does CCO send my data?
No. Configurations, prompts, sessions, paths and keys are not uploaded. Optional anonymous metrics are off by default, and when enabled only a random ID, the month, the version and the agent name go out once a month.
Discussion
No questions or reviews yet.
Related
CLI toolsA scanner that finds leaked keys, dangerous access and connected models in MCP configs of Claude Desktop, Cursor and VS Code.
An authorized bug-bounty suite: 50 subagents, 26 commands and 2 bug-bounty-platform MCP servers for Claude Code, Codex and five more agents
AgentShield audits the .claude folder for plaintext keys, broad permissions, dangerous hooks and risky MCP servers and grades the setup from A to F
MCP-Shield finds your MCP configs, connects to servers and scans tool descriptions for hidden commands, exfiltration channels and tool shadowing
A repository scanner with an MCP server: dependency graph, secrets, vulnerabilities and an API map with code line citations
A security scanner CLI for code and AI agents: vets .claude, hooks, MCP configs and SKILL.md before cloning and finds leaked keys
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Cross-Code Organizer (this card) | no | 387 | |
| MCP Scanner | yes | 1,109 | a Cisco AI Defense scanner that only checks MCP servers, with no settings dashboard or context budget |
| CC Switch | no | 151,986 | a desktop app for switching providers and managing MCP, skills and prompts |
README in short
The README explains that the dashboard began as Claude Code Organizer and grew into a cross-agent tool, and compares it with standalone scanners. Sections follow on Harness Doctor, context budget, the config viewer, the security scanner, the research activation scanner, per-project server control and session distillation. It ends with a table of managed item types, platforms, plans and answers to common questions.