Llm Usage & Cost Tracker
@zhaoyue722
A local-first, multi-provider cost meter for LLM usage, exposed as MCP tools. Captures every call into a local SQLite ledger and lets any coding agent query spend, compare providers, and get recommendations — no cloud, no account. First-class support for Chinese providers (Qwen,
概要
What is Llm Usage & Cost Tracker?
Llm Usage & Cost Tracker is a local-first MCP server and CLI that captures LLM API calls across multiple providers (Anthropic, OpenAI, DeepSeek, Qwen), calculates costs, and answers spend queries through an agent or terminal. It is a cost meter, not a router.
How to use Llm Usage & Cost Tracker?
Install via uv tool install llm-usage-mcp (or clone from source). Set at least one provider API key as an environment variable. Run the capture proxy (llm-usage-proxy), which listens on 127.0.0.1:5525. Point your LLM client’s base URL to the proxy. Use the MCP server (e.g., claude mcp add llm-usage -- uv run llm-usage-mcp) to ask agent questions, or run CLI commands like llm-usage spend, llm-usage compare, and llm-usage recommend.
Key features of Llm Usage & Cost Tracker
- Local-first SQLite storage, no telemetry.
- Supports Anthropic, OpenAI, DeepSeek, and Qwen.
- MCP server with seven query tools.
- CLI with subcommands for spend, compare, recommend.
- Capture proxy logs all API calls idempotently.
- Cross-provider cost comparison and recommendation.
- Works with streaming and non-streaming calls.
- No SaaS signup or account needed.
Use cases of Llm Usage & Cost Tracker
- Monitor daily, weekly, or monthly spend per provider or model.
- Compare hypothetical workload costs across all priced models.
- Get the cheapest model recommendation for a budget and workload.
- Integrate with Claude Code, Cursor, or any MCP client.
- Manually log usage when the capture proxy is not in use.
FAQ from Llm Usage & Cost Tracker
What does it do vs a router?
It is a cost meter, not a router. It tells you what you spent and which provider fits a workload, but never changes your calls. It pairs happily alongside a router or a model-leaderboard tool.
What are the runtime requirements?
Python 3.13+, uv or pipx for installation. SQLite is built into Python, no extra database setup.
Where does my usage data live?
All data is stored locally in a SQLite file at ~/.llm-usage/usage.db.
Which providers are supported?
Anthropic, OpenAI, DeepSeek, and Qwen are supported for both streaming and non-streaming. More providers (Gemini, Bedrock, Moonshot) are planned.
How do I authenticate and connect?
API keys are set as environment variables (e.g., ANTHROPIC_API_KEY). The proxy holds keys server-side; clients point their base URL at http://127.0.0.1:5525 and never need keys directly.
