Llmledger MCP server
io.github.pouriaetab/llmledger·v0.2.0Which AI model, and what it costs: LiteLLM prices, Arena ratings and price history.
0 stars935 downloads/wk
Reviews
Write oneNobody has reviewed Llmledger yet.
If you have run it, two minutes of your experience saves the next person an afternoon.
Llmledger tools (11, 1 write)
write = sends, deletes, buys or postsRead from the package source without running it. The installed server may list more.
compare_modelsCompare 2 to 6 named models for a kind of work: rating, prices, cost per task (or how fast each uses a Claude or ChatGPT plan), the pick, and for each model the share of answer tokens at which it would beat the pick. Use when the user already has candidate models in mind. Not for finding candidates (find_models), for reading a prompt to guess the work (suggest_for_project), or for a monthly budget
estimate_costPrice a workload on 1 to 10 models at each one's cheapest listing: dollars per call, per day and per 30 days, and with cache tokens the saving the cache makes. Use when the question is a budget or a bill: token counts and calls a day are known. Not for deciding which model is better (compare_models) or for pricing past usage from an export (price_my_usage).
estimate_timeEstimate how long one call and a whole job take, as a usual time and a slow day, from the user's own timed runs of that model called that way (kept by log_run). Use when the question is time, not money, and some runs have been logged; it says so when none have. Not for cost (estimate_cost) or for a job you have a promptfoo sample of (forecast_job gives time and cost together).
find_modelsList the highest rated models for a kind of work within price, prompt size, ability and maker limits, each at its cheapest listing with its rating; or, with `name`, look one model up and list every host that offers it. Use when browsing or filtering the catalogue, or to find a model's exact name. Not for choosing between named models (compare_models), for a prompt you have in hand (suggest_for_pro
forecast_jobEstimate a whole job's time, cost and accuracy from a promptfoo run of a sample of it: for each model in the sample, the time as a usual time and a slow day, the cost at its cheapest listing next to the listing the sample used, the 90-day chance of a price change, how often it was right (95% range) and the break-even against the pick. Use when a promptfoo results file of a sample exists and the wh
local_or_cloudCompare running an open model on a computer you buy with paying a cloud model per call: the cost a month each way, the month the computer pays for itself, and whether it keeps up with the load. Use when the question is whether to buy hardware for a local model. Needs timed runs of the local model (log_run) and the computer's price; electricity comes from EIA or the user's own rate. Not for cloud-o
log_runwrite actionRecord a timed run (model, tokens, seconds, how it was called), or every result of a promptfoo results file, in a file on this computer so estimate_time can learn from it. Use right after a call or a promptfoo run whose time is known. Not a query: it stores and returns what it stored. Nothing is sent anywhere.
not_yet_pricedList the models Arena has rated that LiteLLM's catalogue does not price yet, with how soon the catalogue usually prices a maker's models after Arena first shows them. Use when a brand-new model is missing from the other tools' answers, or to ask what is coming. Not for models that already have a price (find_models).
price_historyShow one model's price history: every price and retirement date change since 2023 on every host, each with the LiteLLM change behind it, the chance of a price change, retirement or removal in the next 90 days and year, and with `date`, the prices on that day. Use when the question is about one model over time, a past price, or how likely a price is to move. Not for today's cheapest price of severa
price_my_usagePrice the user's own past coding agent usage, from a ccusage or codeburn export, on every rated model: each session at the price of its day, the models that keep the rating and cost less, a 95% range for the cost per session, and whether two models differ for real or the data cannot tell. Use when a usage export (or a list of token mixes) exists and the question is what the same work would have co
suggest_for_projectSuggest a model and effort level for a prompt or a described project: it guesses the kind of work and names the best value (the cheapest model the votes cannot tell apart from the top), the top rated and the cheapest rated, each with its cost per task. Use when the user asks "which model should I use for this" and the work itself is in hand. Not for named candidates (compare_models) or for browsin
Public scan report
scanner v0.1.10 · 2026-10-26 · same rubric, same numbers if you re-run it
- Code scan11 source files scanned25/25
- –Live reliabilityno gateway calls yet and no remote to proben/a
- –Tool poisoningtools not inspected (local package is not executed); not countedn/a
- Auth qualitylocal package, no credentials required12/15
- Maintenancelast push 3 days ago15/15
- Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10
How Llmledger compares
Against every graded AI and agents MCP server, from our own scans and uptime checks. Updated after each scan run.
- Safety score
- 92/100
- higher than 90% of 4,460 AI and agents servers
- Tools
- 11
- the typical AI and agents server has 6
- Tools that change things
- 9%
- typical: 0%. Fewer write tools means less an agent can break
- Downloads a week
- 935
- more than 92% of 1,538 AI and agents servers with a package
What its tools let an agent do: 10 read · 1 change
Install Llmledger in Claude Code, Cursor or VS Code
claude mcp add llmledger -- uvx llmledger-mcp
What the publisher says
From the Llmledger repository's README, as published. We do not edit it. Read it on GitHub
llmledger
Which AI model should you use, and what will it cost? Paste a prompt and get the best value model for it, the top rated one and the cheapest rated one, with what each costs and the chance its price changes in the next 90 days. Behind it: every price change in LiteLLM's model catalogue since 2023 (4,200+ models), Arena's quality ratings, and a competing risks survival model, rebuilt every day.
Try it: https://pouriaetab.github.io/llmledger/ (no sign up, the prompt stays in your browser)
Use it inside Claude
The same answers inside Claude Code, Claude Desktop, Cursor or any app that speaks MCP. Nothing to host, no keys:
claude mcp add llmledger -- uvx llmledger-mcpEleven tools: suggest a model and effort level for a project, compare models side by side, estimate a budget, find models within limits, a model's price history with its outlook, new models not priced yet, how long a job takes (learned from your own runs), a quote for a whole job from a promptfoo sample, whether buying a computer to run an open model pays off, and what your own coding agent usage (a ccusage or codeburn export) would have cost on every other model. Other apps and details: mcp/README.md.
As an Agent Skill, which stays out of the context until it is used (about 150 tokens against well over 1,000 for the plug-in's tools) and needs only Python:
/plugin marketplace add pouriaetab/llmledger
/plugin install llmledger@llmledgeror npx skills add pouriaetab/llmledger, or copy skills/llmledger into your agent's skills folder.
Watch the models you use
A GitHub Action that fails a check (or just reports) when a model your project uses changes price, gets a retirement date or leaves the catalogue. On the GitHub Marketplace, or add it to a workflow that runs on a schedule:
- uses: pouriaetab/llmledger@main
with:
models: gpt-5, claude-sonnet-4-5, deepseek/deepseek-chat
days: 7 # a price change within this many days counts
fail: "true" # "false" only reports; the report is in the step's outputsKeys are LiteLLM's (gpt-5, not openai/gpt-5; the Action finds the bare key when a prefix is given). The same check from a shell: python action/watch.py --models "gpt-5,claude-sonnet-4-5".
A badge for a model, for any README, refreshed every day:
The address is the key with / as -- and : as - (badge/azure--eu--gpt-5.json). It shows the prices per 1M tokens and the chance of a price change in 90 days (amber from 10%), or the retirement date (orange).
What you get
Shortened. The full README is on GitHub.
Nothing above is checked by us. What we check is on the safety report.
Llmledger: common questions
- Is Llmledger MCP server safe?
- Yes, by our scan: it is graded A (92/100). Read the Llmledger safety report
- How do I install Llmledger?
- It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.
- Does Llmledger need an API key?
- Not as far as the registry entry and our scan can tell: no credentials are declared or required.
- Is Llmledger maintained?
- The last commit was 13 days ago (2026-10-22). The latest release is v0.2.0.
- What can I use instead of Llmledger?
- Servers from other publishers that do the same job: llmtrim MCP server.
- How does Llmledger compare with other AI and agents MCP servers?
- safety score 92/100 (higher than 90% of 4,460 AI and agents servers); tools 11 (the typical AI and agents server has 6); tools that change things 9% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.
Alternatives to Llmledger
Same job from other publishers: the closest match first, then the best rated.
- llmtrimMCP server and proxy that compresses LLM prompts, tool output, and replies to cut token cost.not reviewedEstablishedA
