de en es ja ko ru zh

searxng-mcp

Name: searxng-mcp
Author: 88plug

by 88plug

Overview Schema Source Related Servers Discussions Admin

Python

Hybrid

searxng-mcp

Release Pages

Un servidor MCP para SearXNG que mantiene la salida visible para el modelo breve, preserva los payloads completos de los resultados en metadatos ocultos y admite tanto clientes stdio locales como el uso de streamable-http desplegado.

Qué obtienes

search para una búsqueda web concisa con payloads sin procesar completos en _meta
search_many para fan-out en paralelo, deduplicación y clasificación combinada
search_and_fetch para búsqueda más extracción de fuentes en una sola llamada
research para búsqueda de consultas múltiples con extracciones por lotes y fuentes combinadas
fetch_url para extracción de páginas legibles con citas
fetch_many para extracción de URL en paralelo con almacenamiento en caché
health para el estado del backend, la caché y el renderizado

El servidor está diseñado para ser ligero. SearXNG realiza el trabajo de búsqueda; searxng-mcp maneja el formato de herramientas, el almacenamiento en caché, la extracción y el transporte.

Related MCP server: SearXNG MCP Bridge

Inicio rápido

Instalar desde un checkout

uv sync
uv run searxng-mcp

Instalar con `uvx`

searxng-mcp se distribuye desde este repositorio (sin lanzamiento en PyPI). El camino más corto es:

uvx --from git+https://github.com/88plug/searxng-mcp searxng-mcp

La extracción renderizada se incluye en la instalación predeterminada. Si el host ya tiene Chromium o Chrome, searxng-mcp lo utiliza. De lo contrario, la primera extracción renderizada arranca Playwright Chromium en la caché del usuario automáticamente.

Ejecutar con Docker

docker build -t searxng-mcp .
docker run --rm -p 8811:8811 --add-host=host.docker.internal:host-gateway \
  -e SEARXNG_MCP_BASE_URL=http://host.docker.internal:8890 \
  searxng-mcp

Ejemplo de Claude Desktop

{
  "mcpServers": {
    "searxng-mcp": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/88plug/searxng-mcp", "searxng-mcp"],
      "env": {
        "SEARXNG_MCP_BASE_URL": "http://127.0.0.1:8890",
        "SEARXNG_MCP_TRANSPORT": "stdio"
      }
    }
  }
}

Modos de despliegue

`stdio` local

Úselo para clientes de escritorio y flujos de trabajo privados. Es el modo más simple y seguro.

SEARXNG_MCP_TRANSPORT=stdio uv run searxng-mcp

HTTP transmitible

Úselo para un servicio privado, un despliegue de equipo o un endpoint interno con proxy inverso.

SEARXNG_MCP_TRANSPORT=streamable-http uv run searxng-mcp --host 0.0.0.0 --port 8811

Docker y Compose reforzados

Para un servicio autohospedado de larga duración, utilice la variante de contenedor reforzada y la pila de Compose proporcionada.

docker build -f Dockerfile.prod -t searxng-mcp:prod .
cp docker-compose.env.example .env
docker compose up --build -d

Seguridad y confianza

Este proyecto es seguro para uso local y en redes de confianza. No es un servicio de Internet público listo para usar.

fetch_url y fetch_many pueden obtener URL arbitrarias proporcionadas por el cliente
la extracción renderizada puede iniciar Chromium contra páginas no confiables
streamable-http debe colocarse detrás de una autenticación o un proxy inverso para cualquier despliegue compartido
SEARXNG_MCP_FETCH_VERIFY_TLS=0 es solo para configuraciones de backend privadas o autofirmadas

Si expone el transporte HTTP, trátelo como un servicio interno y añada los controles que esperaría para cualquier herramienta capaz de SSRF.

Por qué existe esto

salida compacta visible para el modelo, con detalles completos preservados en metadatos ocultos
flujos de trabajo de investigación más rápidos mediante búsqueda en paralelo y fan-out de extracción
extracción renderizada para páginas con mucho JS sin banderas de instalación adicionales
despliegue autohospedable para personas que desean su propio servidor MCP respaldado por SearXNG

Superficie MCP

searxng://config expone la configuración actual, el modo de transporte y el soporte de renderizado
searxng://guide resume las herramientas disponibles y cuándo usarlas
quick_lookup, deep_research y research_workflow son prompts de compatibilidad opcionales para clientes que admiten superficies de prompt

Entorno

Variables clave:

SEARXNG_MCP_BASE_URL: URL base de SearXNG, predeterminado http://127.0.0.1:8890
SEARXNG_MCP_FALLBACK_BASE_URLS: instancias de SearXNG de respaldo opcionales separadas por comas
SEARXNG_MCP_TRANSPORT: stdio, streamable-http o sse
SEARXNG_MCP_SEARCH_TIMEOUT: tiempo de espera de búsqueda del backend en segundos
SEARXNG_MCP_FETCH_TIMEOUT: tiempo de espera de extracción en segundos
SEARXNG_MCP_SEARCH_CACHE_TTL: TTL de la caché de búsqueda en segundos
SEARXNG_MCP_FETCH_CACHE_TTL: TTL de la caché de extracción en segundos
SEARXNG_MCP_FETCH_VERIFY_TLS: establecer en 0 para omitir la verificación TLS en las extracciones
SEARXNG_MCP_RENDER_TIMEOUT: tiempo de espera de navegación del navegador para extracciones renderizadas
SEARXNG_MCP_RENDER_WAIT_MS: espera adicional después de la carga del contenido DOM para extracciones renderizadas
SEARXNG_MCP_RENDER_CONCURRENCY: límite de extracción renderizada concurrente
SEARXNG_MCP_RENDER_HEADLESS: establecer en 0 para mostrar el navegador
SEARXNG_MCP_RENDER_BROWSER_PATH: ruta explícita del binario de Chromium o Chrome
SEARXNG_MCP_RENDER_SANDBOX: establecer en 1 para mantener habilitado el sandboxing de Chromium
SEARXNG_MCP_RENDER_BLOCK_RESOURCES: establecer en 0 para permitir imágenes, fuentes, hojas de estilo y medios durante el renderizado
SEARXNG_MCP_RENDER_AUTO_FALLBACK: establecer en 0 para deshabilitar el respaldo renderizado automático
SEARXNG_MCP_RENDER_AUTO_MIN_WORDS: reduzca esto para hacer que el auto-renderizado sea más agresivo
SEARXNG_MCP_RENDER_AUTO_MIN_CHARS: reduzca esto para hacer que el auto-renderizado sea más agresivo
SEARXNG_MCP_CACHE_DIR: ruta del directorio de caché

Benchmarks

uv run searxng-mcp-bench --rounds 3

Los informes de benchmark:

latencia del backend de SearXNG sin procesar
tamaño de salida de búsqueda visible para el token
latencia de búsqueda de consultas múltiples combinadas
latencia de investigación de consultas múltiples
latencia de extracción de búsqueda
latencia de extracción renderizada
latencia de extracción por lotes
latencia de extracción por lotes renderizada

Documentación

Construcción y prueba

uv sync --all-groups
uv run pytest -q
uv run python -m compileall src
uv run mkdocs build --strict

Mantenimiento de dependencias

Verificaciones manuales:

uv lock --upgrade --dry-run
uvx --from pip-audit pip-audit

Verificaciones automatizadas:

PRs semanales de Dependabot para dependencias de uv y GitHub Actions
flujo de trabajo semanal de Dependency Health para escaneo de vulnerabilidades y simulacros de actualización

Available Tools

7 tools

fetch_manyA

Read-onlyIdempotent

Fetch and extract several URLs in parallel. URLs are deduplicated after canonicalisation. Each per-source full text is preserved in hidden _meta.pages[].full_text. Best for: reading a batch of search results, comparing several known sources, building a multi-source citation set. Returns: per-source {title, url, excerpt, citations, render_mode} and aggregate stats (success_count, error_count, elapsed_ms). Rendered fetch is automatic for JS-heavy pages; pass rendered=true to force browser mode for every URL (slower). Use fetch_url for a single URL; research for an end-to-end multi-query workflow.

Name	Required	Description
`urls`	Yes	List of 1–20 absolute http(s) URLs to fetch in parallel. Duplicates are dedupe-d after canonicalisation.
`max_excerpt_chars`	No	Excerpt character cap for the visible output (200–50000). Default comes from server settings. The full extracted text is always returned in hidden `_meta.full_text`.
`max_links`	No	Maximum outbound links to surface in the visible output (0–64, default 8). The complete link list is always included in hidden `_meta`.
`rendered`	No	Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch.
`render_wait_ms`	No	Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly.
`concurrency`	No	Maximum concurrent backend requests for this fan-out (1–16). Higher is faster but puts more load on the SearXNG instance and remote pages. Omit for the server default.
`ttl`	No	Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL.

TDQS

A4.9/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations indicate read-only, idempotent, non-destructive; description adds details like parallel execution, deduplication, full text preservation, and auto-rendering behavior. No contradictions.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Concise and well-structured: starts with purpose, then best-for, returns, parameter details. Every sentence adds value without redundancy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (7 parameters, parallel fetching), the description covers purpose, usage, parameter behavior, and return structure. No output schema but return fields are listed.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, baseline 3. Description adds practical context for each parameter (e.g., when to force rendering, how to set render_wait_ms for SPAs, concurrency load implications), exceeding baseline.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description uses specific verbs ('Fetch and extract several URLs in parallel') and resources, and distinguishes from siblings by mentioning 'fetch_url' for single URL and 'research' for multi-query workflow.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly states best use cases (e.g., reading batch search results, building citation set) and provides alternatives for single URL or multi-query workflows.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

fetch_urlA

Read-onlyIdempotent

Fetch one URL, extract readable content (Readability-style), and return a compact excerpt plus link list. The full extracted text is preserved in hidden _meta.full_text. Best for: reading a known page after search/search_many, extracting clean prose from an article, getting an outbound-link list from a hub page. Returns: {title, url, excerpt, citations, links, content_type, render_mode} plus domain and word/char counts. Rendered fetch is automatic for JS-heavy pages; pass rendered=true to force browser mode when a previous fetch came back nearly empty. Use fetch_many for multiple URLs in one call; search_and_fetch if you do not yet have URLs.

Name	Required	Description
`url`	Yes	Absolute http:// or https:// URL to fetch. Bare domains like 'example.com' are auto-prefixed with https://. Other schemes are rejected.
`max_excerpt_chars`	No	Excerpt character cap for the visible output (200–50000). Default comes from server settings. The full extracted text is always returned in hidden `_meta.full_text`.
`max_links`	No	Maximum outbound links to surface in the visible output (0–64, default 8). The complete link list is always included in hidden `_meta`.
`rendered`	No	Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch.
`render_wait_ms`	No	Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly.
`ttl`	No	Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL.

TDQS

A5/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations provide readOnlyHint, destructiveHint, idempotentHint. Description adds auto-prefixing of bare domains, hidden full text, automatic rendering for JS-heavy pages, and performance trade-offs.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Well-organized: main action, best-for list, return format, parameter details. Every sentence adds value without redundancy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Comprehensive for a tool with 6 parameters and no output schema. Covers return structure, hidden fields, cache, rendering modes, and usage context.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%. Description adds context beyond schema: url scheme handling, default excerpt chars from server, force rendering behavior, and TTL cache override meaning.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Clearly states 'Fetch one URL, extract readable content' and lists return fields. Distinguishes from siblings like fetch_many and search_and_fetch.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly provides 'Best for' scenarios and when to use alternatives. Includes guidance on force rendering when previous fetch returns empty.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

healthA

Read-onlyIdempotent

Report backend, cache, and render-engine readiness. Takes no arguments. Best for: a one-shot check before a session, debugging a 'search failed' result, verifying that the SearXNG backend is reachable and that rendered fetch is available. Returns: {backend, cache, render} with status, latency, and any active fallback URLs.

Name	Required	Description	Default
No parameters

TDQS

A4.5/5.0

Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare readOnlyHint=true, destructiveHint=false, idempotentHint=true. Description adds specific return fields (status, latency, fallback URLs), providing useful behavioral context beyond annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Three sentences: purpose, usage guidance, return value. No wasted words, well front-loaded.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Complete given zero parameters and no output schema. Explains purpose, when to use, and return structure.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

No parameters in schema. Baseline for 0 params is 4. Description mentions 'takes no arguments', confirming no input needed.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Clearly states the tool reports readiness of backend, cache, and render-engine. Distinguishes from sibling tools that take arguments by explicitly noting it takes no arguments.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Provides explicit best-use cases: one-shot check before session, debugging search failures, verifying backend reachability. Does not mention when not to use or alternative tools, but context is clear.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

researchA

Read-only

Run multi-query research: search several queries in parallel, merge and dedupe results, then fetch and extract the strongest sources with citations. Best for: open-ended investigations, building a multi-source briefing, answering a complex question that needs cross-checking across providers. Returns: merged ranking, per-source extracted excerpts and citations, plus a query map showing which queries surfaced each source. Full payloads in hidden _meta. Rendered fetch is automatic for JS-heavy pages; pass rendered=true to force browser mode. Use search_many when extraction is not needed; search_and_fetch for a single query.

Name	Required	Description
`queries`	Yes	List of 1–10 search queries to run in parallel. Use distinct phrasings or angles (synonyms, related entities, opposing framings) for the best merged coverage.
`categories`	No	Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'.
`engines`	No	Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection.
`enabled_engines`	No	Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter.
`disabled_engines`	No	Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter.
`language`	No	BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language.
`pageno`	No	Result page number (1–20, default 1). Use to paginate beyond the first page.
`time_range`	No	Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter.
`safesearch`	No	Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default.
`max_results`	No	Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap.
`fetch_limit`	No	Maximum number of search results to fetch and extract (1–20, default 3). Lower is faster; higher gives more sources at the cost of latency.
`fetch_excerpt_chars`	No	Per-source excerpt character cap (200–20000). Default comes from server settings. Smaller values keep the visible output tight; the full text is always available in hidden `_meta` for each source.
`rendered`	No	Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch.
`render_wait_ms`	No	Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly.
`concurrency`	No	Maximum concurrent backend requests for this fan-out (1–16). Higher is faster but puts more load on the SearXNG instance and remote pages. Omit for the server default.
`ttl`	No	Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL.

TDQS

A4.9/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Discloses key behaviors beyond annotations: parallel queries, merge/dedupe, automatic vs forced rendered fetch, hidden _meta payloads, and parameter effects like concurrency and caching. No contradiction with annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Concise and well-structured: first sentence captures core functionality, followed by best-use cases, return format, render behavior, and alternatives. No redundant sentences.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given 16 parameters and no output schema, the description covers operation, return values (merged ranking, excerpts, citations, query map, hidden _meta), and important nuances (caching, concurrency, render wait). It is contextually complete for effective tool selection.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema has 100% coverage, so baseline is 3. The description adds value by explaining the rendered fetch behavior and defaults (e.g., automatic rendering), but does not detail each parameter. Still, it enhances understanding of key parameters like rendered and max_results.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's purpose: run multi-query research with parallel searches, merging, deduplication, and extraction with citations. It distinguishes from siblings by contrasting with search_many and search_and_fetch.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly provides when to use (open-ended investigations, multi-source briefings) and when not (use search_many if no extraction needed, search_and_fetch for single query), offering clear guidance.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

searchA

Read-only

Search the open web via SearXNG for a single query and return a compact, token-efficient summary of the top results. Best for: quick fact-finding, current events, locating likely sources, single-question lookups. Returns: ranked list of {title, url, domain, snippet, engine, score} plus answers, infoboxes, suggestions and corrections when SearXNG provides them. The full SearXNG payload is also available in hidden _meta.raw_payload. Use search_many instead when running multiple parallel queries; search_and_fetch when you also need extracted page content; research for multi-query investigations with merged sources and fetched excerpts.

Name	Required	Description
`query`	Yes	Search query (1–400 chars). Plain text or SearXNG syntax. SearXNG bangs like '!wikipedia foo' or '!images cats' are supported and route to specific engines.
`categories`	No	Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'.
`engines`	No	Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection.
`enabled_engines`	No	Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter.
`disabled_engines`	No	Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter.
`language`	No	BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language.
`pageno`	No	Result page number (1–20, default 1). Use to paginate beyond the first page.
`time_range`	No	Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter.
`safesearch`	No	Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default.
`max_results`	No	Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap.
`ttl`	No	Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL.

TDQS

A4.7/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Description adds value beyond annotations by explaining the return format (compact summary, raw payload in _meta), without contradicting readOnlyHint=true. Discloses token-efficiency and hidden data, enhancing transparency.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Description is concise (4 sentences), front-loaded with purpose, then usage guidance, then return details and alternatives. Every sentence earns its place.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (11 parameters, no output schema), the description covers purpose, return format, hidden payload, and sibling guidance. Leaves no major gaps for an agent to infer.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema has 100% coverage of parameter descriptions, so baseline is 3. The description does not add additional meaning beyond what the schema provides.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Clearly states the tool searches the open web via SearXNG for a single query and returns a compact summary. Distinct from siblings by naming alternatives explicitly.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Provides a 'Best for' section and explicitly names alternatives (search_many, search_and_fetch, research) with their use cases, offering strong guidance on when to use this tool versus others.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

search_and_fetchA

Read-only

Search SearXNG for one query and fetch+extract the top results in a single call. Combines search and fetch_many for one-query research workflows. Best for: answering one question that needs evidence from full pages, building a citation-backed answer to a single prompt, getting both ranked results and readable excerpts in one round-trip. Returns: search summary plus per-source {title, url, excerpt, citations, render_mode}. The full SearXNG payload and per-page full text are in hidden _meta. Rendered fetch is automatic for JS-heavy pages; pass rendered=true to force browser mode (slower). Use search if extraction is not needed; research for multi-query investigations.

Name	Required	Description
`query`	Yes	Search query (1–400 chars). Plain text or SearXNG syntax. SearXNG bangs like '!wikipedia foo' or '!images cats' are supported and route to specific engines.
`categories`	No	Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'.
`engines`	No	Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection.
`enabled_engines`	No	Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter.
`disabled_engines`	No	Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter.
`language`	No	BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language.
`pageno`	No	Result page number (1–20, default 1). Use to paginate beyond the first page.
`time_range`	No	Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter.
`safesearch`	No	Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default.
`max_results`	No	Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap.
`fetch_limit`	No	Maximum number of search results to fetch and extract (1–20, default 3). Lower is faster; higher gives more sources at the cost of latency.
`fetch_excerpt_chars`	No	Per-source excerpt character cap (200–20000). Default comes from server settings. Smaller values keep the visible output tight; the full text is always available in hidden `_meta` for each source.
`rendered`	No	Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch.
`render_wait_ms`	No	Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly.
`ttl`	No	Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL.

TDQS

A4.4/5.0

Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already indicate read-only and non-destructive behavior. The description adds valuable context about automatic rendered fetch, hidden _meta data, and forced rendering. No contradictions.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is well-structured with clear sections (best for, returns, behavior details). It is somewhat lengthy but each sentence adds value. Slight improvement could be made by condensing the parameter context.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (15 parameters, no output schema, multiple siblings), the description is thorough, covering purpose, usage, return format, hidden data, and behavioral details. It leaves no major gaps.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so baseline is 3. The description does not elaborate on parameters beyond the schema, but it provides context on how the parameters relate to the combined search-and-fetch workflow.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states that the tool combines search and fetch for a single query, and explicitly distinguishes it from sibling tools by recommending 'search' when no extraction is needed and 'research' for multi-query investigations.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides explicit guidance on when to use this tool: 'Best for: answering one question that needs evidence from full pages... getting both ranked results and readable excerpts in one round-trip.' It also clearly states when to use alternatives.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

search_manyA

Read-only

Run several SearXNG searches in parallel, then dedupe and merge their result sets into a single ranked list. Best for: broadening coverage on one topic with synonym/angle variants, comparing how different phrasings rank, building a high-recall source pool before fetching. Returns: merged hits with per-hit queries (which inputs surfaced it), engines, hit_count, and a merged_score. Per-query raw payloads are in hidden _meta. Use search for a single query; research when you also need the top sources fetched and excerpted in the same call.

Name	Required	Description
`queries`	Yes	List of 1–10 search queries to run in parallel. Use distinct phrasings or angles (synonyms, related entities, opposing framings) for the best merged coverage.
`categories`	No	Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'.
`engines`	No	Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection.
`enabled_engines`	No	Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter.
`disabled_engines`	No	Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter.
`language`	No	BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language.
`pageno`	No	Result page number (1–20, default 1). Use to paginate beyond the first page.
`time_range`	No	Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter.
`safesearch`	No	Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default.
`max_results`	No	Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap.
`concurrency`	No	Maximum concurrent backend requests for this fan-out (1–16). Higher is faster but puts more load on the SearXNG instance and remote pages. Omit for the server default.
`ttl`	No	Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL.

TDQS

A4.9/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations indicate read-only, non-destructive, open-world. Description adds valuable behavioral details: deduping, merged scoring, hidden `_meta` for raw payloads, max_results cap behavior, and concurrency/ttl options.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is concise, well-structured with bullet points, and front-loads the core functionality. Every sentence adds value.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Despite 12 parameters and no output schema, the description covers return format (merged hits, per-hit fields, hidden meta) and key behaviors (caching, concurrency). Comprehensive for agent use.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100% so baseline is 3. Description adds value by explaining the meaning of 'queries' (distinct phrasings) and clarifying 'max_results' and 'concurrency' behavior beyond schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states it runs multiple SearXNG searches in parallel, dedupes, and merges results. It distinguishes from siblings by naming 'search' and 'research' as alternatives.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicit 'Best for' scenarios and direct guidance on when to use alternatives: 'Use `search` for a single query; `research` when you also need the top sources fetched and excerpted.'

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

6 tool updatesv0.2.0May 28, 2026

Changedfetch_many15 fields changed

changedInput schema / properties / concurrency / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 16,
+    "minimum": 1,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / concurrency / description

Added value: +"Maximum concurrent backend requests for this fan-out (1–16). Higher is faster but puts more load on the SearXNG instance and remote pages. Omit for the server default."

changedInput schema / properties / max_excerpt_chars / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 50000,
+    "minimum": 200,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / max_excerpt_chars / description

Added value: +"Excerpt character cap for the visible output (200–50000). Default comes from server settings. The full extracted text is always returned in hidden `_meta.full_text`."

addedInput schema / properties / max_links / description

Added value: +"Maximum outbound links to surface in the visible output (0–64, default 8). The complete link list is always included in hidden `_meta`."

addedInput schema / properties / max_links / maximum
```
Added value: +64
```
addedInput schema / properties / max_links / minimum
```
Added value: +0
```

changedInput schema / properties / render_wait_ms / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 15000,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / render_wait_ms / description

Added value: +"Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly."

addedInput schema / properties / rendered / description

Added value: +"Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch."

changedInput schema / properties / ttl / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 86400,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / ttl / description

Added value: +"Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL."

addedInput schema / properties / urls / description

Added value: +"List of 1–20 absolute http(s) URLs to fetch in parallel. Duplicates are dedupe-d after canonicalisation."

addedInput schema / properties / urls / maxItems
```
Added value: +20
```
addedInput schema / properties / urls / minItems
```
Added value: +1
```

Changedfetch_url13 fields changed

changedInput schema / properties / max_excerpt_chars / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 50000,
+    "minimum": 200,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / max_excerpt_chars / description

Added value: +"Excerpt character cap for the visible output (200–50000). Default comes from server settings. The full extracted text is always returned in hidden `_meta.full_text`."

addedInput schema / properties / max_links / description

Added value: +"Maximum outbound links to surface in the visible output (0–64, default 8). The complete link list is always included in hidden `_meta`."

addedInput schema / properties / max_links / maximum
```
Added value: +64
```
addedInput schema / properties / max_links / minimum
```
Added value: +0
```

changedInput schema / properties / render_wait_ms / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 15000,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / render_wait_ms / description

Added value: +"Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly."

addedInput schema / properties / rendered / description

Added value: +"Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch."

changedInput schema / properties / ttl / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 86400,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / ttl / description

Added value: +"Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL."

addedInput schema / properties / url / description

Added value: +"Absolute http:// or https:// URL to fetch. Bare domains like 'example.com' are auto-prefixed with https://. Other schemes are rejected."

addedInput schema / properties / url / maxLength
```
Added value: +4000
```
addedInput schema / properties / url / minLength
```
Added value: +4
```

Changedresearch29 fields changed

addedInput schema / properties / categories / description

Added value: +"Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'."

changedInput schema / properties / concurrency / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 16,
+    "minimum": 1,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / concurrency / description

Added value: +"Maximum concurrent backend requests for this fan-out (1–16). Higher is faster but puts more load on the SearXNG instance and remote pages. Omit for the server default."

addedInput schema / properties / disabled_engines / description

Added value: +"Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter."

addedInput schema / properties / enabled_engines / description

Added value: +"Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter."

addedInput schema / properties / engines / description

Added value: +"Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection."

changedInput schema / properties / fetch_excerpt_chars / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 20000,
+    "minimum": 200,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / fetch_excerpt_chars / description

Added value: +"Per-source excerpt character cap (200–20000). Default comes from server settings. Smaller values keep the visible output tight; the full text is always available in hidden `_meta` for each source."

addedInput schema / properties / fetch_limit / description

Added value: +"Maximum number of search results to fetch and extract (1–20, default 3). Lower is faster; higher gives more sources at the cost of latency."

addedInput schema / properties / fetch_limit / maximum
```
Added value: +20
```
addedInput schema / properties / fetch_limit / minimum
```
Added value: +1
```

addedInput schema / properties / language / description

Added value: +"BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language."

changedInput schema / properties / max_results / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 50,
+    "minimum": 1,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / max_results / description

Added value: +"Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap."

addedInput schema / properties / pageno / description

Added value: +"Result page number (1–20, default 1). Use to paginate beyond the first page."

addedInput schema / properties / pageno / maximum
```
Added value: +20
```
addedInput schema / properties / pageno / minimum
```
Added value: +1
```

addedInput schema / properties / queries / description

Added value: +"List of 1–10 search queries to run in parallel. Use distinct phrasings or angles (synonyms, related entities, opposing framings) for the best merged coverage."

addedInput schema / properties / queries / maxItems
```
Added value: +10
```
addedInput schema / properties / queries / minItems
```
Added value: +1
```

changedInput schema / properties / render_wait_ms / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 15000,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / render_wait_ms / description

Added value: +"Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly."

addedInput schema / properties / rendered / description

Added value: +"Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch."

changedInput schema / properties / safesearch / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      0,
+      1,
+      2
+    ],
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / safesearch / description

Added value: +"Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default."

changedInput schema / properties / time_range / anyOf

Previous value: -[
-  {
-    "type": "string"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      "day",
+      "week",
+      "month",
+      "year"
+    ],
+    "type": "string"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / time_range / description

Added value: +"Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter."

changedInput schema / properties / ttl / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 86400,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / ttl / description

Added value: +"Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL."

Changedsearch19 fields changed

addedInput schema / properties / categories / description

Added value: +"Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'."

addedInput schema / properties / disabled_engines / description

Added value: +"Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter."

addedInput schema / properties / enabled_engines / description

Added value: +"Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter."

addedInput schema / properties / engines / description

Added value: +"Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection."

addedInput schema / properties / language / description

Added value: +"BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language."

changedInput schema / properties / max_results / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 50,
+    "minimum": 1,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / max_results / description

Added value: +"Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap."

addedInput schema / properties / pageno / description

Added value: +"Result page number (1–20, default 1). Use to paginate beyond the first page."

addedInput schema / properties / pageno / maximum
```
Added value: +20
```
addedInput schema / properties / pageno / minimum
```
Added value: +1
```

addedInput schema / properties / query / description

Added value: +"Search query (1–400 chars). Plain text or SearXNG syntax. SearXNG bangs like '!wikipedia foo' or '!images cats' are supported and route to specific engines."

addedInput schema / properties / query / maxLength
```
Added value: +400
```
addedInput schema / properties / query / minLength
```
Added value: +1
```

changedInput schema / properties / safesearch / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      0,
+      1,
+      2
+    ],
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / safesearch / description

Added value: +"Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default."

changedInput schema / properties / time_range / anyOf

Previous value: -[
-  {
-    "type": "string"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      "day",
+      "week",
+      "month",
+      "year"
+    ],
+    "type": "string"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / time_range / description

Added value: +"Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter."

changedInput schema / properties / ttl / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 86400,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / ttl / description

Added value: +"Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL."

Changedsearch_and_fetch27 fields changed

addedInput schema / properties / categories / description

Added value: +"Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'."

addedInput schema / properties / disabled_engines / description

Added value: +"Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter."

addedInput schema / properties / enabled_engines / description

Added value: +"Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter."

addedInput schema / properties / engines / description

Added value: +"Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection."

changedInput schema / properties / fetch_excerpt_chars / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 20000,
+    "minimum": 200,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / fetch_excerpt_chars / description

Added value: +"Per-source excerpt character cap (200–20000). Default comes from server settings. Smaller values keep the visible output tight; the full text is always available in hidden `_meta` for each source."

addedInput schema / properties / fetch_limit / description

Added value: +"Maximum number of search results to fetch and extract (1–20, default 3). Lower is faster; higher gives more sources at the cost of latency."

addedInput schema / properties / fetch_limit / maximum
```
Added value: +20
```
addedInput schema / properties / fetch_limit / minimum
```
Added value: +1
```

addedInput schema / properties / language / description

Added value: +"BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language."

changedInput schema / properties / max_results / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 50,
+    "minimum": 1,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / max_results / description

Added value: +"Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap."

addedInput schema / properties / pageno / description

Added value: +"Result page number (1–20, default 1). Use to paginate beyond the first page."

addedInput schema / properties / pageno / maximum
```
Added value: +20
```
addedInput schema / properties / pageno / minimum
```
Added value: +1
```

addedInput schema / properties / query / description

Added value: +"Search query (1–400 chars). Plain text or SearXNG syntax. SearXNG bangs like '!wikipedia foo' or '!images cats' are supported and route to specific engines."

addedInput schema / properties / query / maxLength
```
Added value: +400
```
addedInput schema / properties / query / minLength
```
Added value: +1
```

changedInput schema / properties / render_wait_ms / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 15000,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / render_wait_ms / description

Added value: +"Extra milliseconds to wait after DOM content load before extracting (0–15000). Use a higher value (e.g. 1500–4000) for SPAs that hydrate slowly."

addedInput schema / properties / rendered / description

Added value: +"Force browser-based rendering for fetches. Default false: the server auto-renders only when a page is JS-heavy or returns near-empty text. Set true when a previous fetch came back nearly empty or to read a known SPA. Forced rendering is several times slower than HTTP fetch."

changedInput schema / properties / safesearch / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      0,
+      1,
+      2
+    ],
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / safesearch / description

Added value: +"Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default."

changedInput schema / properties / time_range / anyOf

Previous value: -[
-  {
-    "type": "string"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      "day",
+      "week",
+      "month",
+      "year"
+    ],
+    "type": "string"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / time_range / description

Added value: +"Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter."

changedInput schema / properties / ttl / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 86400,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / ttl / description

Added value: +"Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL."

Changedsearch_many21 fields changed

addedInput schema / properties / categories / description

Added value: +"Comma-separated SearXNG categories to search (e.g. 'general', 'news', 'images', 'videos', 'science', 'files'). Default is the server's configured category set, typically 'general'."

changedInput schema / properties / concurrency / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 16,
+    "minimum": 1,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / concurrency / description

Added value: +"Maximum concurrent backend requests for this fan-out (1–16). Higher is faster but puts more load on the SearXNG instance and remote pages. Omit for the server default."

addedInput schema / properties / disabled_engines / description

Added value: +"Comma-separated engine names to exclude from this query. Forwarded as SearXNG's 'disabled_engines' parameter."

addedInput schema / properties / enabled_engines / description

Added value: +"Comma-separated engine names to enable in addition to the backend defaults. Forwarded as SearXNG's 'enabled_engines' parameter."

addedInput schema / properties / engines / description

Added value: +"Comma-separated SearXNG engine names to use for this query. Forwarded to SearXNG's 'engines' parameter. Leave unset to use the backend's default engine selection."

addedInput schema / properties / language / description

Added value: +"BCP-47 language hint for SearXNG (e.g. 'en', 'en-US', 'de', 'all'). Default is the server's configured language."

changedInput schema / properties / max_results / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 50,
+    "minimum": 1,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / max_results / description

Added value: +"Maximum visible results to include in the compact summary (1–50). Default comes from server settings (typically 5). The hidden `_meta.raw_payload.results` always contains the full SearXNG response regardless of this cap."

addedInput schema / properties / pageno / description

Added value: +"Result page number (1–20, default 1). Use to paginate beyond the first page."

addedInput schema / properties / pageno / maximum
```
Added value: +20
```
addedInput schema / properties / pageno / minimum
```
Added value: +1
```

addedInput schema / properties / queries / description

Added value: +"List of 1–10 search queries to run in parallel. Use distinct phrasings or angles (synonyms, related entities, opposing framings) for the best merged coverage."

addedInput schema / properties / queries / maxItems
```
Added value: +10
```
addedInput schema / properties / queries / minItems
```
Added value: +1
```

changedInput schema / properties / safesearch / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      0,
+      1,
+      2
+    ],
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / safesearch / description

Added value: +"Safe search level: 0=off, 1=moderate, 2=strict. Omit to use the server default."

changedInput schema / properties / time_range / anyOf

Previous value: -[
-  {
-    "type": "string"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "enum": [
+      "day",
+      "week",
+      "month",
+      "year"
+    ],
+    "type": "string"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / time_range / description

Added value: +"Restrict results to recent content. Valid values: 'day', 'week', 'month', 'year'. Omit for no time filter."

changedInput schema / properties / ttl / anyOf

Previous value: -[
-  {
-    "type": "integer"
-  },
-  {
-    "type": "null"
-  }
-]New value: +[
+  {
+    "maximum": 86400,
+    "minimum": 0,
+    "type": "integer"
+  },
+  {
+    "type": "null"
+  }
+]

addedInput schema / properties / ttl / description

Added value: +"Cache TTL override in seconds (0–86400). 0 disables caching for this call. Omit to use the server's default TTL."

7 tool updatesv0.1.0Apr 15, 2026
- First observedfetch_many
- First observedfetch_url
- First observedhealth
- First observedresearch
- First observedsearch
- First observedsearch_and_fetch
- First observedsearch_many

Deploy Server

Let your AI agent see the webWeb-to-screenshot MCP for Claude, Cursor and Codex. No cookie banners. Free to start.snap-render.comAd

TDQS

A4.6/5.0

Scored 2026-05-28 07:40 across 7 tools

Disambiguation4/5

Tools have distinct purposes, but search_and_fetch and research are similar enough that an agent might misselect if descriptions aren't carefully read. However, descriptions clearly differentiate them.

Naming Consistency4/5

Most tools follow a verb_noun pattern in snake_case (fetch_url, search_many), but 'health' and 'research' deviate slightly. Overall pattern is recognizable.

Tool Count5/5

Seven tools cover the core operations of searching, fetching, and health checking without unnecessary bloat. The count is appropriate for the server's purpose.

Completeness5/5

The tool surface covers all typical workflows: single/multi search, single/multi fetch, combined search+fetch, and multi-query research. Health check ensures backend status. No obvious gaps.

Maintenance

ActivitySlowing

ResponsivenessNo issues

Resources

Related MCP Connectors

Novada MCPOAuth
io.github.NovadaLabs
One MCP server for all web data. Scrape, search, crawl, and AI research in one install.
Browser Automation Web Scraping Search
13
https://mcp.novada.com/mcp
Keenable Web Search A
ai.keenable.api
Docs: https://docs.keenable.ai/mcp-server Keenable is a free, remote MCP server that gives agents access to the web index. Search the web with ranked results and date/site filters, then fetch any indexed page as clean markdown. Works out of the box with no account or API key.
Research & Data Search Web Scraping
23
https://api.keenable.ai/mcp?keenable_title=glama
Serper A
io.github.pipeworx-io
Serper MCP — wraps the Serper Google Search API (serper.dev)
Search RAG Systems
472
https://gateway.pipeworx.io/serper/mcp
mcp A
ai.llmse
Public MCP server for the LLM Search Engine
Search RAG Systems Developer Tools
102
https://llmse.ai/mcp

Related MCP Servers

searxng
Search Agent Orchestration
SecretiveShell
A
license
B
quality
F
maintenance
An MCP server for connecting agentic systems to search systems via searXNG.
Updated 5 months ago (2026-05-29 20:54 UTC)
1
131
MIT
SearXNG MCP Bridge
Search Browser Automation Web Scraping
nitish-raj
A
license
Not graded
quality
F
maintenance
An MCP server that acts as a bridge to SearXNG instances, allowing AI models to perform web searches with privacy-focused search tools. It supports both STDIO and HTTP transports, featuring built-in rate limiting and session management.
Updated 2 months ago (2026-08-25 19:11 UTC)
33 npm
7
MIT
searxng-mcp
Search Browser Automation
fighter3005
F
license
A
quality
D
maintenance
MCP server that provides a search_web tool to query a self-hosted SearXNG instance and return structured web search results.
Updated 4 months ago (2026-06-29 14:46 UTC)
1
-
Basic MCP Tools
Browser Automation Search
Tyguy047
F
license
Not graded
quality
C
maintenance
A self-hosted MCP server providing private web search, web page fetching, and current date/time tools, powered by a bundled SearXNG instance for API-key-free local search.
Updated 4 months ago (2026-07-08 22:08 UTC)
2
-

searxng-mcp

Qué obtienes

Inicio rápido

Instalar desde un checkout

Instalar con uvx

Ejecutar con Docker

Ejemplo de Claude Desktop

Modos de despliegue

stdio local

HTTP transmitible

Docker y Compose reforzados

Seguridad y confianza

Por qué existe esto

Superficie MCP

Entorno

Benchmarks

Documentación

Construcción y prueba

Mantenimiento de dependencias

Available Tools

Tool Schema Changelog

Maintenance

Resources

Related MCP Connectors

Related MCP Servers

Instalar con `uvx`

`stdio` local