AnySearch MCP Server

English | 简体中文

Unified real-time search MCP server supporting general web search, vertical domain search, parallel batch search, and full-page URL content extraction.

Features

• General Web Search — open-ended natural language queries

• Vertical Domain Search — structured queries across finance, academic, security, legal, code, and more

• Parallel Batch Search — execute multiple independent queries in one call

• URL Content Extraction — fetch and extract full page content as Markdown

• Anonymous Access — works without an API key (with lower rate limits)

Related MCP server: Search MCP Server

API Key Configuration

An API key is optional but recommended. Without a key, all features still work via anonymous access with lower rate limits.

Register for an API Key (Recommended)

The agent can register the user and obtain an API key in a single call — no verification code, no manual signup. Ask the user for a real email address: it becomes the account username.

curl -s -X POST "https://api.anysearch.com/v1/auth/email/register" \
  -H "Content-Type: application/json" \
  -d '{"email": "you@example.com"}'

Success response (code: 0) returns the account info and a one-time plaintext API key:

{
  "code": 0,
  "message": "success",
  "data": {
    "username": "you@example.com",
    "email": "you@example.com",
    "login_url": "https://www.anysearch.com/login",
    "api_key": {
      "id": "key_xxxxxxxx",
      "key": "as_sk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
      "key_prefix": "as_sk_xxxxxx...",
      "name": "default",
      "rate_limit": 100,
      "quota_limit": 0,
      "expires_at": null,
      "created_at": "2026-06-23T10:23:00Z"
    }
  }
}

Quota fields such as rate_limit are account-specific response values; do not treat the example as a fixed plan.

On success the agent MUST:

1. Put data.api_key.key into the MCP server config as the Authorization: Bearer <key> value (and/or set ANYSEARCH_API_KEY) — it is shown only once (it can also be retrieved later from the dashboard).

2. Tell the user their username (= email) and the login_url.

Errors can be expressed through the HTTP status, application-level code, or message; do not assume every error uses code: -1:

The email must be real and reachable.

Get an API Key

Alternatively, visit https://anysearch.com/console/api-keys to create a free API key.

Key Priority

Key Behavior

MCP Transport

The production endpoint is:

https://api.anysearch.com/mcp

It natively uses Streamable HTTP. Current OpenCode, Claude Code, Cursor, VS Code, Cline, OpenAI Codex, Google Antigravity, DeepSeek Harness, Hermes Agent, and OMP releases can connect to it directly; no SSE or stdio proxy is needed. The configurations below follow each client's current official documentation.

Installation

The API key is optional. Where a client documents environment-variable or secret storage for custom headers, the examples use the recommended authenticated setup. Where it does not, the example stays anonymous and explains how to add authentication only in private user configuration. To use anonymous access elsewhere, remove only the Authorization entry and keep X-Anysearch-Client.

OpenCode

Official docs: MCP servers and configuration locations.

Use ~/.config/opencode/opencode.json for a global configuration or opencode.json in the project root. On Windows, the global path is %USERPROFILE%\.config\opencode\opencode.json.

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "anysearch": {
      "type": "remote",
      "url": "https://api.anysearch.com/mcp",
      "enabled": true,
      "oauth": false,
      "headers": {
        "Authorization": "Bearer {env:ANYSEARCH_API_KEY}",
        "X-Anysearch-Client": "mcp/1.0.0"
      }
    }
  }
}

OpenCode uses {env:NAME} for environment-variable substitution. "oauth": false prevents an unnecessary OAuth discovery flow for this API-key-authenticated server.

After saving, run opencode mcp list and confirm that anysearch is shown as connected.

Claude Code

Official docs: Connect Claude Code to tools via MCP.

For a private, user-wide installation, run:

claude mcp add --transport http anysearch https://api.anysearch.com/mcp --scope user --header 'Authorization: Bearer ${ANYSEARCH_API_KEY}' --header 'X-Anysearch-Client: mcp/1.0.0'

The user scope stores the server in ~/.claude.json and makes it available across all projects on the machine. Single quotes preserve the variable reference so the plaintext key is not written to the config; set ANYSEARCH_API_KEY before starting Claude Code. For anonymous access, omit the Authorization --header option.

For a shareable project configuration, create .mcp.json in the project root:

{
  "mcpServers": {
    "anysearch": {
      "type": "http",
      "url": "https://api.anysearch.com/mcp",
      "headers": {
        "Authorization": "Bearer ${ANYSEARCH_API_KEY}",
        "X-Anysearch-Client": "mcp/1.0.0"
      }
    }
  }
}

Claude Code expands ${VAR} and ${VAR:-default} in .mcp.json, including inside url and headers. Set ANYSEARCH_API_KEY before starting Claude Code. Project-scoped servers require approval when first opened interactively.

Verify the connection with:

claude mcp get anysearch
claude mcp list

Inside Claude Code, /mcp shows the server status and available tools.

Cursor

Official docs: Model Context Protocol.

Use .cursor/mcp.json for a project or ~/.cursor/mcp.json globally:

{
  "mcpServers": {
    "anysearch": {
      "url": "https://api.anysearch.com/mcp",
      "headers": {
        "Authorization": "Bearer ${env:ANYSEARCH_API_KEY}",
        "X-Anysearch-Client": "mcp/1.0.0"
      }
    }
  }
}

Cursor automatically recognizes the remote HTTP endpoint and supports ${env:NAME} interpolation in url and headers.

Check the connection under Customize > MCPs or Output > MCP Logs. If Cursor CLI is installed, you can also run:

agent mcp list
agent mcp list-tools anysearch

VS Code

Official docs: Add and manage MCP servers and MCP configuration reference.

Run MCP: Open User Configuration for a user-wide server, or create .vscode/mcp.json in a workspace. This example uses a password input so the key is prompted for and stored securely instead of being committed:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "anysearch-api-key",
      "description": "AnySearch API key",
      "password": true
    }
  ],
  "servers": {
    "anysearch": {
      "type": "http",
      "url": "https://api.anysearch.com/mcp",
      "headers": {
        "Authorization": "Bearer ${input:anysearch-api-key}",
        "X-Anysearch-Client": "mcp/1.0.0"
      }
    }
  }
}

For anonymous access, remove the Authorization entry and the entire inputs array.

Run MCP: List Servers, select anysearch, then choose Start or Show Output to verify the connection. Note that ${input:...} works in local VS Code, but servers requiring interactive input are not forwarded to Agent Host. For configurations shared with Agent Host or other Copilot tools, use a workspace .mcp.json or user ~/.copilot/mcp-config.json and avoid interactive inputs.

Cline

Official docs: MCP.

In the Cline panel, open MCP Servers > Configure > Configure MCP Servers, or add a hosted endpoint from the Remote Servers tab. Use the explicit streamableHttp type; omitting it falls back to legacy SSE behavior.

{
  "mcpServers": {
    "anysearch": {
      "type": "streamableHttp",
      "url": "https://api.anysearch.com/mcp",
      "headers": {
        "Authorization": "Bearer ${env:ANYSEARCH_API_KEY}",
        "X-Anysearch-Client": "mcp/1.0.0"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Cline IDE supports ${env:VAR} in url, headers, and env. Run cline mcp to open the CLI wizard, or cline config mcp --json to list the configuration. The official web docs still name ~/.cline/mcp.json, while the current source defaults to ~/.cline/data/settings/cline_mcp_settings.json; use Configure MCP Servers in the IDE or the CLI wizard to open the effective file instead of relying on a hard-coded path.

The following five entries were checked against first-party configuration documentation on 2026-08-27. This establishes documented client compatibility, not a local end-to-end connection certification.

OpenAI Codex

Official docs: Model Context Protocol.

Use ~/.codex/config.toml globally or .codex/config.toml in a trusted project. The Codex CLI, IDE extension, and desktop app share this configuration:

[mcp_servers.anysearch]
url = "https://api.anysearch.com/mcp"
bearer_token_env_var = "ANYSEARCH_API_KEY"
http_headers = { "X-Anysearch-Client" = "mcp/1.0.0" }

Set ANYSEARCH_API_KEY before starting Codex. For anonymous access, remove bearer_token_env_var. Run codex mcp list, or use /mcp inside the Codex TUI, to confirm the server and its tools are available.

Google Antigravity

Official docs: MCP servers in Antigravity.

Use ~/.gemini/config/mcp_config.json globally or .agents/mcp_config.json in a workspace. Remote servers must use serverUrl; the legacy url and httpUrl keys are not supported:

{
  "mcpServers": {
    "anysearch": {
      "serverUrl": "https://api.anysearch.com/mcp",
      "headers": {
        "X-Anysearch-Client": "mcp/1.0.0"
      }
    }
  }
}

Antigravity's official documentation currently shows literal custom-header values but does not document environment-variable interpolation for remote headers. If authentication is required, add "Authorization": "Bearer YOUR_API_KEY" only to the private global file and never commit it to .agents/mcp_config.json. Open MCP Servers > Manage MCP Servers in the IDE, or run /mcp in the CLI, to inspect live status, logs, and tools.

DeepSeek Harness

Official sources: DeepSeek Harness and @deepseek-ai/dsh-mcp-client.

DeepSeek Harness configures MCP through its first-party Cordis plugin rather than a standalone mcp.json. Merge this patch into ~/.dsh/profiles/web/cordis.patch.yml (or the corresponding profile under $DSH_HOME):

- insert:
    - id: mcp-anysearch
      name: "@deepseek-ai/dsh-mcp-client"
      config:
        serverName: anysearch
        transport: streamable-http
        url: "https://api.anysearch.com/mcp"
        headers:
          Authorization: !!js '`Bearer ${process.env.ANYSEARCH_API_KEY}`'
          X-Anysearch-Client: "mcp/1.0.0"
        failOnStartupError: true

Set ANYSEARCH_API_KEY, then run npx @deepseek-ai/dsh --profile web --dump-config to verify that the profile includes mcp-anysearch before starting it with npx @deepseek-ai/dsh web. The current MCP plugin exposes tools only; MCP resources and prompts are not yet supported. DeepSeek Harness is a developer preview, so recheck the linked plugin documentation when upgrading.

Hermes Agent

Official docs: MCP integration.

Add the server to $HERMES_HOME/config.yaml and store ANYSEARCH_API_KEY in $HERMES_HOME/.env or the process environment. On macOS and Linux, $HERMES_HOME defaults to ~/.hermes; the native Windows installer currently defaults to %LOCALAPPDATA%\hermes:

mcp_servers:
  anysearch:
    url: "https://api.anysearch.com/mcp"
    headers:
      Authorization: "Bearer ${ANYSEARCH_API_KEY}"
      X-Anysearch-Client: "mcp/1.0.0"

Verify it with hermes mcp test anysearch and hermes mcp list. After editing the configuration during a session, run /reload-mcp or start a new session.

OMP (Oh My Pi)

Official docs: MCP configuration.

Use .omp/mcp.json in a project or ~/.omp/agent/mcp.json globally:

{
  "$schema": "https://raw.githubusercontent.com/can1357/oh-my-pi/main/packages/coding-agent/src/config/mcp-schema.json",
  "mcpServers": {
    "anysearch": {
      "type": "http",
      "url": "https://api.anysearch.com/mcp",
      "headers": {
        "Authorization": "Bearer ${ANYSEARCH_API_KEY}",
        "X-Anysearch-Client": "mcp/1.0.0"
      }
    }
  }
}

OMP expands ${VAR} and ${VAR:-default} in configuration values. Set ANYSEARCH_API_KEY, then run /mcp reload, /mcp test anysearch, and /mcp list inside OMP.

Client Quick Reference

Available Tools

search

Execute a search query — general or vertical domain.

get_sub_domains

Query the vertical domain directory. Required before any search that uses a domain — returns valid sub_domains and their parameter schemas.

Returns a Markdown table: sub_domain | description | params

batch_search

Execute 1–5 independent search queries in parallel. Single failure does not block others.

extract

Fetch full page content from a URL and return as Markdown. Truncated at 50,000 characters. HTML pages only.