io.github.pouriaetab/llmledger

1 of 11 tools are described as changing something (creating, editing, deleting, sending or running).

• find_modelson 7 other serversreads or searches

  “List 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 h…”

  Publisher declares: read-only yes · idempotent yes · open-world no

  7 inputs

  • taskstring“The kind of work, as Arena rates it”
  • max_output_pricenumber or null“Highest answer price, dollars per 1M tokens”
  • min_contextinteger or null“Smallest prompt size the model must take, in tokens”
  • needsarray“What the model must be able to do”
  • makerstring or null“Only this maker, for example anthropic or openai”
  • limitinteger“How many models to list (up to 200 for the whole catalogue)”
  • namestring or null“Look up one model by name instead of browsing: all its listings”
• compare_modelson 35 other serversreads or searches

  “Compare 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 wou…”

  Publisher declares: read-only yes · idempotent yes · open-world no

  6 inputs, 1 required

  • modelsarrayrequired“2 to 6 model names, for example claude-opus-5.5, gpt-6.1-sol”
  • taskstring“The kind of work, as Arena rates it”
  • input_tokensinteger“Prompt tokens per task”
  • output_tokensinteger“Answer tokens per task, including hidden thinking”
  • answer_tokensobject or null“Answer tokens per model when they differ, by the names given”
  • paystring“api: dollars; claude or chatgpt: how fast each uses that plan”
• estimate_coston 48 other servers

  “Price 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.”

  Publisher declares: read-only yes · idempotent yes · open-world no

  6 inputs, 3 required

  • modelsarrayrequired“1 to 10 model names”
  • input_tokensintegerrequired“Prompt tokens per call”
  • output_tokensintegerrequired“Answer tokens per call, including hidden thinking”
  • cached_tokensinteger“How many of the prompt tokens are read from the cache”
  • calls_per_daynumber“Calls per day”
  • cache_write_tokensinteger“How many of the prompt tokens are written to the cache”
• price_historyon 33 other serversreads or searches

  “Show 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 nex…”

  Publisher declares: read-only yes · idempotent yes · open-world no

  2 inputs, 1 required

  • modelstringrequired“A model name, for example claude-opus-5.5, or one listing such as bedrock/anthropic.claude-opus-5-5”
  • datestring or null“A day as YYYY-MM-DD: each listing's prices that day and the change that set them”
• suggest_for_projecton 1 other server

  “Suggest 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…”

  Publisher declares: read-only yes · idempotent yes · open-world no

  3 inputs, 1 required

  • promptstringrequired“The project's starting prompt, or a description of the work”
  • paystring“api: dollars; claude or chatgpt: compare within that plan”
  • answerstring“Typical answer length”
• not_yet_pricedreads or searches

  “List 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.”

  Publisher declares: read-only yes · idempotent yes · open-world no

  1 input

  • makerstring or null“Only this maker, for example meta or ibm”
• estimate_time

  “Estimate 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).”

  Publisher declares: read-only yes · idempotent yes · open-world no

  7 inputs, 3 required

  • modelstringrequired“A model name, as the catalogue or Ollama names it”
  • input_tokensintegerrequired“Prompt tokens per call”
  • output_tokensintegerrequired“Answer tokens per call, including hidden thinking”
  • callsinteger“How many calls the job makes”
  • waystring“How the model is called”
  • thinkingboolean“Thinking turned on”
  • concurrencyinteger“Calls running at once”
• log_runcreates or changes datadescribed as changing data

  “Record 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.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world no

  8 inputs

  • modelstring or null“The model that ran”
  • input_tokensinteger or null“Prompt tokens of the call”
  • output_tokensinteger or null“Answer tokens of the call, hidden thinking included”
  • secondsnumber or null“How long the call took”
  • waystring or null“How the model was called (default api; for a promptfoo file, read from each provider)”
  • thinkingboolean“Thinking was on (a promptfoo result that counts thinking tokens is marked on by itself)”
  • taskstring or null“The kind of work, if known”
  • promptfoo_resultsstring or null“Path of a promptfoo results file (promptfoo eval -o results.json): every result becomes a run”
• forecast_job

  “Estimate 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 …”

  Publisher declares: read-only yes · idempotent yes · open-world no

  6 inputs, 2 required

  • promptfoo_resultsstringrequired“Path of a promptfoo results file from a sample of the job (promptfoo eval --filter-sample 20 -o sample.json)”
  • itemsintegerrequired“How many tests the whole job has”
  • concurrencyinteger or null“Calls running at once in the whole job (default: as the sample ran)”
  • waystring or null“How the models are called (default: read from each provider)”
  • thinkingboolean“Thinking turned on”
  • wrong_answer_costnumber or null“Dollars a wrong answer costs you to catch and fix; the pick then counts it”
• local_or_cloud

  “Compare 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.”

  Publisher declares: read-only yes · idempotent yes · open-world no

  13 inputs, 6 required

  • local_modelstringrequired“The open model you would run at home, as Ollama names it, for example llama3.1:8b”
  • cloud_modelsarrayrequired“1 to 5 cloud models to compare with”
  • input_tokensintegerrequired“Prompt tokens per call”
  • output_tokensintegerrequired“Answer tokens per call, including hidden thinking”
  • calls_per_monthnumberrequired“Calls a month”
  • hardware_pricenumberrequired“What the computer costs you, in dollars”
  • wattsnumber or null“Power the computer draws while it answers, in watts”
  • electricity_pricenumber or null“Your electricity price, dollars per kWh (default: EIA's average for homes)”
  • statestring or null“A US state, for EIA's average there, for example Texas”
  • monthsinteger“How many months to compare over”
  • cached_tokensinteger“Prompt tokens read from the cloud model's cache”
  • local_waystring“How the local model is called”
  • thinkingboolean“Thinking turned on for the local model”
• price_my_usage

  “Price 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…”

  Publisher declares: read-only yes · idempotent yes · open-world no

  5 inputs

  • exportstring or null“Path of a token export on this computer: `ccusage daily --json > f.json`, `ccusage session --json > f.json`, or `codeburn export -f json -o f.json`”
  • mixarray or null“Instead of a file: token mixes as {model, date, input, output, cache_read, cache_write, session?}”
  • taskstring or null“The kind of work the sessions were (default: codeburn's categories, else coding)”
  • comparearray or null“Two models to test against each other on the same sessions”
  • limitinteger“How many models to list in every_model (the cheapest first)”

Each quoted sentence is the first sentence of the publisher’s own description. The kind of tool and the “changes data” mark are our AI model’s reading of that description, not something we tested: two differently worded runs had to agree, otherwise no label is shown, and nobody has checked them by hand. A tool without the “changes data” mark is described as only reading, searching or calculating, or the two runs did not agree. No tool was called. Read the publisher’s full description before relying on it. “Publisher declares” lists the hints the server itself sends with the tool (MCP annotations: read-only, destructive, idempotent, open-world); they are the publisher's claims, not checked by us. Inputs are the parameters in the tool's own input schema: name, declared type, whether it is required, and the first sentence of its description. Read 2026-10-23.

The registry verifies a package once, when it is published, and does not re-check it. A difference below is about the listing, not a judgement of the publisher.