{"id":"perplexityai-modelcontextprotocol","name":"modelcontextprotocol","homepage":"https://docs.perplexity.ai/guides/mcp-server","repo_url":"https://github.com/perplexityai/modelcontextprotocol","category":"ai-ml","subcategories":[],"tags":["mcp","model-context-protocol","perplexity","web-search","research","reasoning","typescript","stdio","http","api-integration"],"what_it_does":"Provides an MCP server that exposes Perplexity API capabilities as MCP tools for AI assistants, enabling real-time web search and model-based Q&A/research (search, ask, research, reasoning). Supports both stdio (typical MCP clients) and an optional HTTP deployment mode.","use_cases":["Web search with ranked results for up-to-date information","Answering questions with integrated web search (sonar-pro)","Producing deep research reports (sonar-deep-research)","Complex reasoning and problem-solving (sonar-reasoning-pro)","Embedding Perplexity capabilities into MCP-compatible clients (Cursor, Claude Desktop, VS Code, etc.)"],"not_for":["Use cases requiring fully offline operation","Applications that need fine-grained, per-end-user auth/accounting enforced by the MCP server itself (only an API key is described)","Scenarios where you cannot manage outbound network access (including proxy/firewall constraints)"],"best_when":"You want to add real-time web-grounded capabilities to an MCP-capable assistant using the documented Perplexity API key configuration.","avoid_when":"You need strict, documented rate-limit semantics or idempotency guarantees at the MCP layer (not specified).","alternatives":["Use Perplexity APIs directly from your own backend (instead of via MCP)","Use other MCP servers for search/research (different providers)","Build a lightweight MCP wrapper around Perplexity’s Search/Chat endpoints"],"af_score":66.8,"security_score":63.5,"reliability_score":37.5,"package_type":"mcp_server","discovery_source":["github"],"priority":"high","status":"evaluated","version_evaluated":null,"last_evaluated":"2026-03-30T13:21:10.234067+00:00","interface":{"has_rest_api":true,"has_graphql":false,"has_grpc":false,"has_mcp_server":true,"mcp_server_url":null,"has_sdk":true,"sdk_languages":["TypeScript","JavaScript"],"openapi_spec_url":null,"webhooks":false},"auth":{"methods":["API key via PERPLEXITY_API_KEY"],"oauth":false,"scopes":false,"notes":"Authentication is described as a single Perplexity API key. No OAuth flow or scope model is documented for the MCP server."},"pricing":{"model":null,"free_tier_exists":false,"free_tier_limits":null,"paid_tiers":[],"requires_credit_card":false,"estimated_workload_costs":null,"notes":"Pricing for Perplexity usage is not described in the provided content; MCP server just forwards requests to Perplexity models/APIs."},"requirements":{"requires_signup":true,"requires_credit_card":false,"domain_verification":false,"data_residency":[],"compliance":[],"min_contract":null},"agent_readiness":{"af_score":66.8,"security_score":63.5,"reliability_score":37.5,"mcp_server_quality":85.0,"documentation_accuracy":80.0,"error_message_quality":0.0,"error_message_notes":null,"auth_complexity":95.0,"rate_limit_clarity":20.0,"tls_enforcement":90.0,"auth_strength":70.0,"scope_granularity":20.0,"dependency_hygiene":60.0,"secret_handling":75.0,"security_notes":"Security observations are based only on provided README/manifest data. TLS is implied via HTTPS default base URL, but explicit enforcement details (e.g., redirect/HTTP rejection) are not documented. Auth is an API key without described fine-grained scopes. README advises using environment variables for configuration and includes an optional debug log level (DEBUG may increase risk of leaking sensitive context if misused). Proxy configuration support exists, which can introduce additional security considerations (credential-in-proxy URL).","uptime_documented":0.0,"version_stability":60.0,"breaking_changes_history":50.0,"error_recovery":40.0,"idempotency_support":"false","idempotency_notes":null,"pagination_style":"none","retry_guidance_documented":false,"known_agent_gotchas":["Some MCP clients may fail due to npx installation/stdout noise; README suggests using `npx -yq` to suppress output.","Long research queries may require increasing `PERPLEXITY_TIMEOUT_MS` (default 5 minutes)."]}}