Flux Search API: pricing, use cases and tools on Dreamtune

Flux Search and Flux Lineage are available through Dreamtune's tools catalog. Search discovers pages on the web. Lineage selects passages from a page or document that answer your question, with source positions and context for names and dates.

This guide explains what each tool does, Flux's published costs, and how to call both through Dreamtune's API and TypeScript SDK. Pricing was checked against Flux's live English pricing page on October 16, 2026.

What is Flux Search?

Flux is a search and retrieval API for AI agents, language-model applications and retrieval-augmented generation, or RAG. Its two core tools solve different parts of gathering evidence.

Flux Search accepts a query and returns ranked web results with titles, URLs and snippets. You can ask it to apply Lineage to some or all results. The enrich option controls how many pages it reads for supporting passages; enrich: 0 requests search results without that page-reading step.

Flux Lineage accepts document text or a URL. It selects passages relevant to your questions and returns both the original passage and enriched text. That enriched text can resolve a pronoun to a person's name or a relative date to an explicit date, based on the source. The response retains source spans and annotations so your application can inspect the supporting material.

These descriptions follow the official Flux API reference. A returned passage supports checking an answer; it does not establish that every statement on the underlying page is correct.

How much does Flux Search cost?

Flux's published pricing lists $4 per 1,000 searches, equivalent to $0.004 per search.

At that published rate, 100 searches cost $0.40 and 10,000 searches cost $40. These are search-only calculations, not estimates of an entire agent workflow. Model inference and other tools have their own costs.

Search responses include the billed search count, the number of results returned, how many results received Lineage enrichment, and the actual reported cost. Use that receipt to measure a real workflow rather than inferring cost from the number of links in its answer.

How much does Flux Lineage cost?

The live English pricing page lists $0.10 per million input tokens for Lineage, with output tokens free as checked on October 16, 2026.

For example, processing 50,000 input tokens for one question costs $0.005 at that rate. Five questions over that same document represent five passes, or 250,000 billed input tokens, costing $0.025. Flux documents that each question is a separate pass over its document.

Published prices can change. Flux's per-request usage.cost_usd_ticks field is the integer cost authority, with usage.cost_usd supplying the corresponding decimal display value. Dreamtune converts that verified cost to its USD micro-unit ledger, rounding up at that boundary, and applies the workspace's configured usage fee. The completed Dreamtune run shows its actual cost, fee and total. Provider pricing is separate from any wallet top-up fee.

When to use Search and when to use Lineage

Use Search when your agent needs to discover sources: product research, documentation discovery, company research, or finding recent reporting. Set a result count appropriate to the task and inspect publication dates when freshness matters.

Use Lineage when you already have the document or page: extracting evidence from an article, selecting relevant documentation, or preparing focused context for a model. It is especially useful when short snippets would lose the identity behind “she” or the date behind “last Tuesday.”

Use both when the task needs discovery followed by evidence extraction. Keep the question with the passage, retain the source URL and span, and let the reader inspect the original material. Measure source relevance and citation support on your own tasks before drawing accuracy or cost-saving conclusions.

Where are the Flux tools available in Dreamtune?

Find Flux Search and Flux Lineage by searching for “Flux” in the public catalog or the workspace Tools page. Each listing opens the Playground with the corresponding tool selected.

Developers use provider fluxsearch and endpoint /api/v1/search or /api/v1/lineage. Both use Dreamtune's existing tool-run API, SDK, CLI and MCP tools. A Dreamtune platform key needs the appropriate tool scopes, and execution requires available workspace credit. Dreamtune holds the upstream credential; you do not include a Flux key in the request body.

Call Flux Search through the Dreamtune API

Submit a tool run to POST https://dreamtune-api.fly.dev/v1/developer/tool-runs with your Dreamtune bearer key and a unique Idempotency-Key header. This is the JSON body:

{
  "provider": "fluxsearch",
  "endpoint": "/api/v1/search",
  "input": {
    "body": {
      "query": "Flux Search API documentation",
      "max_results": 3,
      "enrich": 1
    }
  }
}

The run returns its state, identifier and completed output when available. Read or refresh that saved run to check its result. Reuse an idempotency key only for the identical operation; a fresh key represents a new potentially billable search.

Use Flux Lineage with the TypeScript SDK

Use the existing @dreamtune/sdk client. Keep your platform key in your application's server-side secret configuration.

import { DreamtuneClient } from "@dreamtune/sdk";

const client = new DreamtuneClient({
  apiKey: process.env.DREAMTUNE_API_KEY!,
  baseUrl: "https://dreamtune-api.fly.dev",
});

const run = await client.createToolRun({
  provider: "fluxsearch",
  endpoint: "/api/v1/lineage",
  input: {
    body: {
      url: "https://fluxsearch.io/en/pricing",
      queries: ["How much does Flux Search cost?"],
    },
  },
  idempotencyKey: crypto.randomUUID(),
});

For discovery, call client.listTools({ q: "fluxsearch" }). The CLI exposes the same catalog through dreamtune tools list --q fluxsearch, and Dreamtune's MCP list_tools accepts that query too. See the Dreamtune documentation for authentication and tool-run methods.

What should an agent keep from the response?

Keep the source URLs, original passages, enriched passages and span information that the task needs. Preserve Flux's enrichment status: a successful search can contain fewer enriched pages than requested, and its usage fields describe that shortfall. Avoid presenting an unenriched snippet as a verified quotation.

Track completed cost, latency, useful sources and citation support across representative tasks. That gives you a practical basis for choosing between a quick search, a focused document extraction and a longer research workflow.

Browse Flux tools on Dreamtune, or read our guide to building reliable tool-calling applications.