
modelcost
io.github.CloudCorpRecords/modelcost
AI API Radar: cheapest LLM models, per-model price history, provider status and price changes.
Documentation
modelcost
Which LLM API is cheapest right now, what did a model cost last month, and is its provider up? modelcost is a local forwarding client for the hosted AI API Radar MCP service: it answers tools/list locally from the captured tool definitions and forwards every call to the service, returning the answers exactly as the service sends them back.
Three of the tools are pure reads over public pricing and status data: a cheapest-model ranking by blended per-token cost, a dated price history for a single model, and live provider operational status collected from public status pages. The fourth, price_change_alerts, is dual-mode: called without a webhook it reads recent price changes, and called with one it subscribes a webhook URL you control to price-change events. modelcost itself sends nothing anywhere — it forwards the arguments you give it and originates no messages.
All 4 captured tools are built; there are no omissions.
The tools
cheapest_model— Rank the cheapest LLM API models that meet a constraint, by blended $/M-token cost. Filter by minimum context window (min_context) andprovider; weight input vs output price within_weight/out_weight;operational_onlydrops providers with a major or critical status. No required parameters.price_history— The dated input + output $/M-token price history for one model, useful for spotting price-war cuts and trends.modelis required.status— Current operational status for one provider (passprovider) or for all of them (omit it), from each provider's public status page. No required parameters.price_change_alerts— Read recent price changes, or subscribe your own webhook to future ones. Passwebhook(an HTTPS URL) to subscribe, optionally filtered byprovideranddirection(cutorraise); omit it to read. No required parameters.
Install and run
As an MCP server via npm:
npx -y @reneturcios/modelcost mcp
MCP client configuration:
{
"mcpServers": {
"modelcost": {
"command": "npx",
"args": ["-y", "@reneturcios/modelcost", "mcp"]
}
}
}
As a container:
docker run -i ghcr.io/cloudcorprecords/modelcost
docker run --rm ghcr.io/cloudcorprecords/modelcost status '{}'
CLI
One binary serves both the command line and the MCP server. Each tool is a subcommand taking a single JSON string with that tool's arguments:
modelcost cheapest_model '{"min_context":128000,"limit":5}'
modelcost price_history '{"model":"anthropic--claude-opus-4-x"}'
modelcost status '{}'
modelcost price_change_alerts '{"direction":"cut"}'
modelcost --version
Exit codes: 0 when the forwarded call succeeds, 1 when the upstream result is an error, and 2 for usage problems such as an unknown subcommand or a malformed JSON argument.
Configuration
AIRADAR_MCP_URL— Overrides the upstream endpoint. The default is the pinned hosted endpoint,https://aiapi.dropwatchhq.com/mcp. (The override variable is a reconstruction added by this rebuild.)
The upstream registry record declares no headers, so no authentication header is ever sent; the tool surface was captured keyless. A webhook URL passed to price_change_alerts is registered with the service exactly as given — it should be an endpoint you control.
Upstream failures — a JSON-RPC error from the service, a non-2xx response, an unreadable body, or a request that fails outright — are reported to the caller as an error result whose text starts with modelcost error:.
How forwarding works
The first forwarded call opens a session: an initialize request (protocol version 2025-03-26), followed by a notifications/initialized notification, then the tools/call itself. If the service answers the initialize request with an Mcp-Session-Id header, that value is carried on every later request, and the open session is reused for the rest of the process. Replies are accepted either as plain JSON or as server-sent events. Successful results pass through verbatim, including any isError flag and any extra fields the service includes.
Development
npm install
npm run build # tsc -> dist/ (dist/cli.js is the bin)
npm test # tsx --test test/*.test.ts — stubbed upstream, no network
npm run dev # run the CLI from source
The test suite checks the served tool list against the captured surface (all 4 tools, deep-equal, in order, with no titles invented and no required key added to the three tools that carry none), exercises the session handshake and its reuse, verifies that no Authorization header is ever sent under any environment, and verifies that arguments go upstream untouched and results come back untouched.
License
MIT — see LICENSE. Copyright (c) 2026 Rene Turcios.
@reneturcios/modelcostnpm install @reneturcios/modelcostghcr.io/cloudcorprecords/modelcost:0.1.0docker pull ghcr.io/cloudcorprecords/modelcost:0.1.0:undefinedRelated Servers
io.github.CloudCorpRecords/abandoncheck
Judge whether a dependency is maintained, from factual signals: releases, commits, archival.
io.github.CloudCorpRecords/actionbridge
Zapier exposed actions as MCP tools via a generic action runner.
io.github.CloudCorpRecords/actionpin
GitHub Actions supply-chain hardener: audit workflows for unpinned actions and build a SHA pin plan.