@calcis/langchain

Calcis cost-estimation callback for LangChain.

Live pricing for 25+ models, side-by-side comparisons, and a web estimator: https://calcis.dev

• Full price index: https://calcis.dev/models
• Compare models: https://calcis.dev/compare
• API reference: https://calcis.dev/api-docs

Attach the handler to any LangChain runnable and get a one-line cost estimate for every LLM call, plus a rolling session total.

Install

npm install @calcis/langchain @langchain/core

You also need a Calcis API key (Pro tier or above). Get one at calcis.dev/dashboard.

Usage

import { ChatAnthropic } from "@langchain/anthropic";
import { CalcisCallbackHandler } from "@calcis/langchain";

const calcis = new CalcisCallbackHandler({
  apiKey: process.env.CALCIS_API_KEY!,
});

const model = new ChatAnthropic({
  model: "claude-sonnet-4-6",
  callbacks: [calcis],
});

const response = await model.invoke("Explain quantum computing");

console.log(`Session total: $${calcis.getSessionTotal().toFixed(4)}`);

Every LLM call now emits:

[calcis] claude-sonnet-4-6 · 243 in · 650 out · $0.0105 · session: $0.0105 (1 calls)

Config

new CalcisCallbackHandler({
  apiKey: string,                // required: calc_… key
  verbose?: boolean,             // default true: logs per-call summary
  onEstimate?: (e) => void,      // optional structured sink
});

The estimate object passed to onEstimate:

{
  model: string,
  inputTokens: number,
  outputTokens: number,
  inputCost: number,
  outputCost: number,
  totalCost: number,
  confidence: "high" | "medium" | "low"
}

Session helpers

calcis.getSessionTotal(); // cumulative cost across calls
calcis.getCallCount();    // number of LLM calls so far
calcis.resetSession();    // zero the counters (e.g. per request)

Failure mode

This handler never throws. If the Calcis API is unreachable, returns an error, or the response is malformed, the handler silently skips the estimate for that call: the underlying LangChain chain keeps running. Cost estimation is a nice-to-have; your product is not.

Model name mapping

LangChain's provider integrations don't always name models the way Calcis does. The handler ships a small translation map:

| LangChain | Calcis | | ----------------- | ----------------- | | claude-3-5-sonnet | claude-sonnet-4-6 | | claude-3-opus | claude-opus-4-6 | | gpt-5.4 | gpt-5-4 |

Unmapped names are passed through unchanged. Use calcis models (or visit calcis.dev/models) for the full list of recognised IDs.

Links

• Web App
• API Docs
• CLI (calcis)
• Pricing dataset (@calcis/pricing)

License

MIT