by ccarter01011
Ask Claude, ChatGPT, Claude Code, Cursor or VS Code who spent your AI budget and what to fix. TraxAgent's read-only MCP server and Claude Code plugin.
# Add to your Claude Code skills
git clone https://github.com/ccarter01011/traxagent-claude-pluginGuides for using ai agents skills like traxagent-claude-plugin.
See how traxagent-claude-plugin compares with popular alternatives.
traxagent-claude-plugin is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by ccarter01011. Ask Claude, ChatGPT, Claude Code, Cursor or VS Code who spent your AI budget and what to fix. TraxAgent's read-only MCP server and Claude Code plugin. It has 88 GitHub stars.
traxagent-claude-plugin's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/ccarter01011/traxagent-claude-plugin" and add it to your Claude Code skills directory (see the Installation section above).
traxagent-claude-plugin is primarily written in Python. It is open-source under ccarter01011 on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh traxagent-claude-plugin against similar tools.
No comments yet. Be the first to share your thoughts!
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM lists this project for discovery and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
See comparison
Who spent your AI budget last month, and on what? Ask Claude, ChatGPT, Claude Code, Cursor, Gemini or VS Code, and get the answer from your real usage.

Two questions, answered from the sample company: what each team spent last week, then which cost cuts to make first. The second answer ranks the fixes by effort, not just by size, so the safe config changes come before the ones that need a quality test.
Your teams may pay for AI in 5 places or more: the Anthropic, OpenAI and Gemini APIs, Claude Code or Codex, Cursor, and Claude or ChatGPT seats. Each vendor's console is correct about its own slice and blind to the rest.
TraxAgent pulls them into one place every night, by team and by person. Then an AI review goes through your usage and tells you where you're overpaying. Each fix comes with what it costs you per month and the exact change to make. You walk into the budget meeting with the full number, where it went, and what you've already done to bring it down.
This repo is the Claude Code plugin. The same read-only server, https://www.traxagent.io/mcp, also works in Claude, ChatGPT, Cursor, VS Code and any MCP client: see Use it in other AI tools.
After installing (below):
/traxagent:spend-audit shows what your own Claude Code use costs at API prices, by model, project, day and session, plus your cache hit rate and what one model tier down would save. Claude reads Claude Code's local logs with its own tools and sends nothing anywhere.ai-spend-demo server answers from sample data with no sign-in, so you can see every tool before connecting your own.Behind the plugin is a full dashboard. The answers come from TraxAgent, which also has a dashboard for your team: spend by team and person, fixes ranked in dollars, budgets and caps, an AI registry, and finance reports. Click around the demo, no sign-up.
Example answer:
AI spend, Nov 10–16: $1,842.47 total (about $263.21/day, list prices)
| Team | Spend |
|----------------------|---------|
| Backend Team | $126.73 |
| FrontEnd Team | $98.10 |
| Data Science | $214.06 |
| Viper Team | $231.67 |
| Atlas Team | $11.13 |
| APIs (Prod) | $431.28 |
| APIs (other) | $218.42 |
| AWS Bedrock Seats | $511.08 |
Biggest open cost fix: none right now.
And the savings question, on the sample company:
You: Using TraxAgent: what cost cuts do you recommend across our AI spend?
$9.2k–18.7k a month of $39.0k is avoidable. Ranked by effort, not size:
1. Batch API for 2 nightly Data jobs $2.8k–4.0k runs at 01:00 UTC, no one waits
2. Newer versions: Sonnet 4.6 → 5.5 $1.9k–3.4k cheaper per token, same family
3. Opus → Sonnet for CI code review $772 85% vs 86% success, half the cost
4. Prompt caching on 16 keys $5.1k–14.1k hit rate under 30% on most
Do the first two this week: config changes with no quality risk.
Caching is the big prize, and one key's cache broke on Nov 6.
| Ask | What you get |
|---|---|
| "What cost cuts do you recommend across our AI spend?" | Every fix grouped by kind and ranked by effort, with the monthly range and the safe ones first |
| "Which fix should we ship first, and can you do it here?" | The steps for the top fix; Claude makes the change in this repo and opens a pull request with the saving |
| "Is Opus worth it for our code reviews, or would Sonnet do?" | Success rate and cost per success for each model on that task, from your own outcomes |
| "Why did AI spend jump on Tuesday?" | The key, model and team behind the jump, and whether a cache broke or a loop ran away |
| "Who are our top 10 AI spenders this month, across Claude Code, Codex and Cursor?" | Spend per person with their team and a split by tool |
| "Which AI tools have no owner or aren't approved?" | The AI registry: every key, coding agent, seat plan and app in use, with what needs a look |
| "Which API keys haven't been used in 30 days?" | Keys to revoke, from the registry's last-used dates |
| "If we moved this traffic to Sonnet 5.5, what would it cost?" | Last month's real usage replayed at the other model's list prices |
In Claude Code:
/plugin marketplace add ccarter01011/traxagent-claude-plugin
/plugin install traxagent@traxagent
Then run /mcp, pick plugin:traxagent:ai-spend, choose Authenticate and sign in to TraxAgent in your browser. There's no token to copy. The MCP server is on TraxAgent's paid plans.
If this saved you a budget meeting, a ⭐ on this repo helps other teams find it.
No account yet? Set one up in about 5 minutes by pasting a read-only admin key. TraxAgent is free up to $100 a month of AI spend.
Every option below signs in with your TraxAgent account (OAuth). Access is read-only and there's no token to copy.
| Tool | How to add TraxAgent |
|---|---|
| Claude (web, desktop) | Open TraxAgent in Claude's connector directory, choose Connect, then sign in. |
| ChatGPT | Plugins → Add → Add custom MCP server. URL: https://www.traxagent.io/mcp, Authentication: OAuth. Create it, sign in, then mention @TraxAgent in a chat. |
| Cursor | Add to Cursor, then sign in when Cursor asks. Also on Cursor Directory. |
| VS Code / GitHub Copilot | Add to VS Code, sign in, then use it from Copilot chat in agent mode. |
| Any MCP client (Codex, Gemini CLI, Windsurf…) | Add https://www.traxagent.io/mcp as a remote (streamable HTTP) MCP server. Clients that support OAuth sign in. Others can send a read-only API token from API and MCP. |
It's also listed in the official MCP Registry as io.traxagent/traxagent.
Some companies now cap each engineer's AI spend by day, week or month. If leadership asks whether your teams need one, test a cap against your own numbers before you set it. TraxAgent shows who would hit it and how often, then warns people early, before they reach it, and again at 80% and 100%.