Agent Search MCP
by mFurukawa10
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MODE | No | Runtime mode, for example MODE=http to run the server in HTTP mode. | |
| PROXY_URL | No | Explicit proxy URL used together with USE_PROXY=true. | |
| USE_PROXY | No | Set to true together with PROXY_URL to use an explicit proxy. | |
| EXA_API_KEY | No | Optional API key for Exa Search, a neural search provider (en, zh). | |
| YDC_API_KEY | No | Optional API key for You.com Search, a commercial Web Search provider (en, zh). | |
| OUTPUT_STYLE | No | Set to compact to reduce response tokens. Compact mode keeps full detail for the first results and reduces later entries to source-preserving references. | |
| RERANK_TOP_K | No | Top-K used by optional semantic reranking. | |
| BOCHA_API_KEY | No | Optional API key for Bocha Web Search, a Chinese-first AI Search provider (zh, en). | |
| BRAVE_API_KEY | No | Optional API key for Brave Search, a commercial Web Search provider (en, zh). | |
| ENABLED_TOOLS | No | Tool registration allowlist. Deny wins over allow. | all |
| DENIED_ENGINES | No | Engine execution denylist. Deny wins over allow. | none |
| DISABLED_TOOLS | No | Tool registration denylist. Deny wins over allow. | none |
| SEMANTIC_DEDUP | No | Enables optional semantic deduplication. Semantic processing is the only optional feature that uses Python and Model2Vec. | |
| SERPER_API_KEY | No | Optional API key for Serper Google Search, a Google SERP search provider (en, zh, auto). | |
| SNIPPET_LENGTH | No | Controls snippet length to reduce response tokens. | |
| TAVILY_API_KEY | No | Optional API key for Tavily Search, an agent-oriented search provider (en, zh). | |
| WIBY_PROXY_URL | No | Explicit proxy URL for the Wiby adapter. Engine-specific single-proxy variables take precedence over their pool. | |
| ALLOWED_ENGINES | No | Engine execution allowlist. Deny wins over allow. | all |
| ALLOWED_ORIGINS | No | Browser requests with an Origin header must match ALLOWED_ORIGINS. | |
| DEDUP_THRESHOLD | No | Threshold used by optional semantic deduplication. | |
| HTTP_AUTH_TOKEN | No | Authentication token for HTTP mode. HTTP mode requires HTTP_AUTH_TOKEN unless you set HTTP_ALLOW_UNAUTHENTICATED=true. | |
| SEMANTIC_RERANK | No | Enables optional semantic reranking. Semantic processing is the only optional feature that uses Python and Model2Vec. | |
| SOGOU_PROXY_URL | No | Explicit proxy URL for the Sogou adapter. Engine-specific single-proxy variables take precedence over their pool. | |
| WIBY_PROXY_URLS | No | User-owned proxy pool for Wiby as a JSON array of 2-16 HTTP(S) proxy URLs. | |
| MAX_FULL_RESULTS | No | Number of results that keep full detail when compact output is used. | |
| MOJEEK_PROXY_URL | No | Explicit proxy URL for the Mojeek adapter. Engine-specific single-proxy variables take precedence over their pool. | |
| SOGOU_PROXY_URLS | No | User-owned proxy pool for Sogou as a JSON array of 2-16 HTTP(S) proxy URLs. | |
| MOJEEK_PROXY_URLS | No | User-owned proxy pool for Mojeek as a JSON array of 2-16 HTTP(S) proxy URLs. | |
| PAID_ENGINE_ORDER | No | Selects the first configured optional provider; not a quality claim. | brave,exa,tavily,youcom,tencent_wsa,bocha,serper |
| SEARCH_CACHE_TTL_MS | No | Time-to-live for the exact-result cache. | |
| TENCENT_WSA_API_KEY | No | Optional API key for the Tencent Web Search API, an official Chinese Web Search provider (zh). | |
| DUCKDUCKGO_PROXY_URL | No | Explicit proxy URL for the DuckDuckGo adapter. Engine-specific single-proxy variables take precedence over their pool. | |
| SEARCH_PROVIDER_MODE | No | Default routing policy: free_first, quality_escalation, paid_first, or free_only. The default free_first policy never spends a configured API credential; free_only blocks paid providers. | free_first |
| DUCKDUCKGO_PROXY_URLS | No | User-owned proxy pool for DuckDuckGo as a JSON array of 2-16 HTTP(S) proxy URLs. | |
| EVIDENCE_BUDGET_CHARS | No | Evidence-character budget, caps query-relevant passages across the complete response. | 1200 |
| SEARCH_CACHE_DIRECTORY | No | Opts into local persistence of the exact-result cache. The default exact-result cache stays in memory. | |
| SEARCH_BUDGET_MAX_CALLS | No | Adapter-attempt budget. | 16 |
| SEARCH_CACHE_MAX_ENTRIES | No | Maximum number of entries in the exact-result cache. | |
| SEARCH_BUDGET_MAX_RESULTS | No | Admitted raw-result budget. | 100 |
| HTTP_ALLOW_UNAUTHENTICATED | No | Set to true to allow running HTTP mode without HTTP_AUTH_TOKEN. | |
| SEARCH_BUDGET_MAX_ELAPSED_MS | No | End-to-end elapsed-time budget. | 30000 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Server capabilities have not been inspected yet.
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
This server cannot be deployed
Maintenance
ActivityActive
ResponsivenessNo issues