AI search API with cited answers your agents can verify

    An AI search API lets your agent send one natural-language prompt and get back cited sources, with an optional written answer. Desearch AI Search runs POST /desearch/ai/search across the web and X and returns titles, links, snippets and highlights. Usage is $0.40 per 1000 items, and new accounts get $5 in API credits after attaching a card in Console.

    Get your API keySearch API guide
    $0.40/1000 items
    usage rate
    Web and X
    select per request
    MCP + REST
    agent-ready access
    Live answer flow
    POST /desearch/ai/search
    Agent query

    What patterns make production AI agents more reliable?

    Sources found
    1

    How cited answers are assembled

    example.com

    2

    Reading a search snippet

    docs.example.com

    3

    Highlights from a source page

    research.example.org

    Cited answer

    Reliable agents pair focused tools with fresh retrieval, cited sources, and outputs downstream systems can inspect. 123

    citations
    freshness
    MCP-ready setup
    cost metadata

    Response shape

    What the AI search API returns

    POST /desearch/ai/search returns the sources it found, and a written answer only when you ask for one.

    search

    An array of web results. Each item has a title, a link, a snippet, and highlights.

    tweets

    Included when tools contains twitter. These are the X posts for the same prompt.

    completion

    An optional written answer. It is returned when result_type is LINKS_WITH_FINAL_SUMMARY.

    Direct integration

    Call the AI search API: request and response

    Prefer direct API control?

    Send the same Search endpoint from your application. The curl example calls POST https://api.desearch.ai/desearch/ai/search with Authorization and Content-Type headers.

    POST /desearch/ai/search reference

    POST /desearch/ai/search
    curl --request POST \
  'https://api.desearch.ai/desearch/ai/search' \
  --header 'Authorization: $DESEARCH_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "prompt": "latest research on AI search benchmarks",
    "tools": ["web", "twitter"],
    "streaming": false,
    "count": 10
  }'

    Request parameters

    prompt
    Required. The natural-language question.
    tools
    Required. web, twitter, or both.
    start_date, end_date
    Optional UTC timestamps, YYYY-MM-DDTHH:MM:SSZ. start_date and end_date apply to X results.
    date_filter
    Optional relative window, such as PAST_WEEK. On AI Search, date filtering applies to X results.
    streaming
    Optional boolean. Defaults to true. Set false for one JSON response.
    result_type
    ONLY_LINKS or LINKS_WITH_FINAL_SUMMARY. Defaults to LINKS_WITH_FINAL_SUMMARY.
    system_message
    Optional text that guides the written answer.
    scoring_system_message
    Optional text used when scoring results.
    count
    Optional integer. Results per source. Minimum 10, maximum 200, default 10.

    Response fields

    search
    Web results with title, link, snippet, and highlights.
    tweets
    X posts, present when twitter is in tools.
    completion
    Optional written answer when result_type is LINKS_WITH_FINAL_SUMMARY.
    warnings
    Non-fatal notes when a requested tool returns no results.

    Billing headers

    • X-Desearch-Cost-Usd
    • X-Desearch-Usage-Count
    • X-Desearch-Service
    • X-Desearch-Currency

    Trusted by AI teams

    View case studies

    Agent-ready integration

    Set up the AI search API with one prompt (MCP)

    Paste the onboarding prompt into Claude Code, Codex, Cursor, VS Code, or any agent that can read instructions. It connects Desearch through MCP and proves the setup with a real request.

    Agent onboarding prompt

    Use curl to read www.desearch.ai/agents.md and follow it to set up Desearch

    One instruction replaces a setup guide: your agent reads the source of truth, installs the connection, and verifies that Search is ready.

    Agent setup
    waiting for view
    Search verifiedMCP connected

    Your agent can now search fresh sources and return evidence it can cite.

    Built for agent workflows

    AI search API use cases for agents

    Connect one search capability, then let each agent use the evidence in the way its task requires.

    01

    Answer with current context

    Give assistants a direct answer plus the links and passages users need to check it.

    • Support
    • Copilots
    • Chat interfaces
    02

    Research with traceable sources

    Turn a broad question into ranked evidence and a concise summary without a custom search stack.

    • Market research
    • Monitoring
    • Briefings
    03

    Ground autonomous workflows

    Give planning and monitoring agents fresh evidence they can cite before they recommend or take the next step.

    • Research agents
    • Monitoring agents
    • Decision support

    Source control

    Choose sources: web and X

    Supported tools: web, twitter. Other values are unsupported and may be rejected (for example, youtube returns 422). To focus web results on specific sites, keep web and add include_domains.

    For links without a written answer, call POST /desearch/ai/search/links/web or POST /desearch/ai/search/links/twitter. See links only or cited answers in one request for a worked example.

    What builders say

    Selected excerpts from builders using Desearch for research, trading, and agent workflows. Lightly edited for readability.

    View all

    ExponentialReed

    @ExponentialReed

    I just connected mine to Bittensor TAO subnet 22, Desearch, with an API endpoint for X data to help me with subnet analysis.

    The results are getting where they should be.

    53912.9K views

    CryptoZPunisher

    @CryptoZPunisher

    SN22 is becoming an intelligence layer for agents. And behind agents come predictions, then portfolio management.

    Revenue and aligned incentives. The market probably hasn't fully priced this yet.

    43103,886 views

    Numinous AI

    @numinous_ai

    The forecasting environments just got richer.

    Starting this Thursday, miners can choose to take on the cost of news retrieval via Desearch.

    53165,849 views

    vaNlabs

    @vaNlabs

    Desearch integration complete.

    crustty now has full social sentiment guidance, and it's integrated (weighted) in his subnet valuations. All thanks to Bittensor SN22, Desearch.

    2853,537 views

    Angry Dave

    @AngryDavee

    Desearch AI is somewhat like Google but decentralized and smarter.

    AI-powered search to deliver smart, summarized results. APIs for developers to integrate decentralized search.

    206770 views

    Nuance Subnet

    @NuanceSubnet

    How SN22 Desearch helps incentivize a more nuanced discourse on X

    Desearch has a powerful API for searching through X posts, which has been crucial for our subnet in order to find, verify and measure engagement on post on X.

    Explore Desearch: https://desearch.ai

    27171,713 views

    AI search API FAQ

    What a request returns, which sources you can search, streaming, and usage.

    Search API guide

    An AI search API takes a natural-language prompt, searches live sources, and returns the sources it found, with an optional generated answer that cites them. Desearch does this with POST /desearch/ai/search across the web and X.

    search is an array of items with title, link, snippet, and highlights. tweets is included for X when tools contains twitter. completion is an optional written answer when result_type is LINKS_WITH_FINAL_SUMMARY. Dates are request filters: start_date and end_date apply to X results.

    To search Reddit, use the Web Search API with site:reddit.com (see Reddit Search API)

    Yes. Set streaming to true to receive the generated summary through server-sent events. Keep it false when your application prefers one complete JSON response.

    AI Search is usage-based at $0.40 per 1000 items. Each request is billed for the count you request (10 to 200), whichever sources you select and however many results come back. Requests that fail with 503 or a 422 validation error are not billed. New accounts get $5 in API credits after attaching a card in Console. Usage is pay-as-you-go after that. Successful billable responses expose request cost and usage metadata, and Desearch Console shows the current rate, balance, and request history.

    The API uses standard HTTP status codes. Check the current rate-limit guide for your account, handle 429 responses with backoff, and use the API reference for the complete request and response schemas.

    Ready when your agent is

    Connect your agent to live, cited search.

    Start through the agent workflow, then use the API directly whenever your application needs finer control.

    Get your API keyPOST /desearch/ai/search reference
    Earn up to $155 in API credits
    1. 01

      Copy one prompt

      Paste the agents.md instruction into the coding agent you already use.

    2. 02

      Let it connect

      Your agent installs Desearch MCP and configures the Search tool.

    3. 03

      Review the check

      Setup finishes only after the key and a cited Search request are verified.