Desearch

Install

Works with Claude Code, VS Code, Codex, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/desearch-ai-mcp-desearch

About

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list, install config and code. Checked Oct 26, 2026.

Tools

20 read · 0 write · 2 delete

Secrets

2 (API keys, tokens or passwords)

Install scripts

None found

Runs as

Node.js package

Tools (22)

As the server listed them on Oct 25, 2026. Descriptions are the server’s own words; hidden characters are shown as [U+…].

• ai-searchAI Searchread-only

  AI search and analysis on web using Desearch AI

• extractExtract Page Contentread-only

  Extract a public URL and return its content as plain text or HTML using Desearch. Preferred over web-crawl for new integrations. API docs: https://www.desearch.ai/docs/api-reference/get-web-extract

• fetchFetch Pageread-only

  Read a public page as text. Pass a URL returned by search.

• people-profilesList People Foundread-only

  Everyone your People Searches found on one platform, one entry per person: their best-scoring result and the search that found them (found_by). Free; it does not run a new search. Search by name, handle, bio, or location with q, filter by contact channels (has), audience, fit score, location, or verification, and sort by score, recent, or audience. People from searches that ended more than 30 days ago are left out.

• people-searchPeople Searchcan change or delete data

  Start a paid job billed to the user's Desearch balance: find people on X, YouTube, LinkedIn, or TikTok who match a plain-language description. Each person comes back with a fit score, why they matched, their audience, and public contacts and links. This call reserves count times the platform's per-person price (reserved_cost_usd) and returns a task_id at once. When the search ends you pay only for the people it delivered (a per-search minimum applies when it finds anyone; a search that finds no one is free) and the rest of the reservation is returned. The search runs in the background, usually…

• people-search-controlControl People Searchcan change or delete data

  Pause, resume, or cancel a People Search task. Free. pause: a queued search pauses at once; a running one finishes its current batch first (status pausing, then paused). A search paused for more than 7 days is cancelled. resume: continues a paused search where it stopped, with no new charge. cancel: stops the search for good; you pay only for the people it already delivered and the rest of the reservation is returned. People found so far stay readable for 30 days.

• people-search-statusGet People Search Statusread-only

  Status, progress, and cost of one People Search task, plus a page of the people found so far: name, handle, url, platform, audience, location, fit score, why they matched, contacts, and links. Free. People arrive while the search runs: pass the returned next_cursor as cursor on the next call to get only the people found since then. When done is true and has_more is false you have everyone the search found. An ended task has completion_reason (why it stopped), coverage_limited: true when a budget stopped it before count (more people may match), and warnings such as partial_enrichment (some prof…

• people-search-tasksList People Search Tasksread-only

  Your People Search tasks, newest first, with status, counts, cost, why each ended search stopped (completion_reason, coverage_limited), and the metadata labels set on people-search. Free. Use it to find a task_id (filter by metadata to find the searches of one campaign), or to see which searches are still queued, running, or paused: a new search is refused while too many are active.

• searchSearchread-only

  Search the public web and return titles and links. Use fetch with a result URL to read the page.

• web-crawlCrawl Web Page (Legacy)read-only

  Crawl a public URL and return its content as plain text or HTML on the legacy Desearch /web/crawl route. The SDK marks webCrawl deprecated in favor of extract; this tool stays for parity with that route. Prefer extract for new integrations. API docs: https://www.desearch.ai/docs/guide/apis/extract-api

• web-links-searchWeb Links Searchread-only

  Search the web for links using Desearch. Only the web source is accepted.

• web-searchWeb Searchread-only

  SERP-style web search using Desearch. Returns ranked titles, links, and snippets.

• x-links-searchX Links Searchread-only

  AI search for X (Twitter) post links using Desearch. Returns links from posts that match the prompt.

• x-post-by-idGet X Post by IDread-only

  Fetch a single X (Twitter) post by its ID.

• x-post-repliesGet X Post Repliesread-only

  Fetch replies to an X (Twitter) post, with an optional keyword query.

Signals

Stars

1

Forks

1

Open issues

2

Last commit

Oct 22, 2026 (7 days ago)

Last release

Oct 21, 2026 (8 days ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name

io.github.Desearch-ai/mcp-desearch

Version

0.3.3

Auth

API key

Packages

npm desearch-mcp-server

License

MIT

Repository

github.com/Desearch-ai/mcp-desearch

Website

www.desearch.ai/docs/guide/sdk/mcp

First seen

Oct 24, 2026

Last checked

Oct 25, 2026

Source

Official MCP Registrygithub-graphql