Tavily alternatives

Exa provides a web search API designed for machine consumption: embeddings-based retrieval, full page content extraction, similarity search from a URL and structured research endpoints.

You.com started as a consumer AI search engine and now focuses on selling web search, news and research APIs that give agents current, citable data.

Deepgram offers speech-to-text, text-to-speech and a combined voice agent API designed for real-time conversational systems where latency matters.

Together AI serves open-weight models through a fast inference API and also rents dedicated GPU clusters for training.

Fireworks AI provides optimised inference for open-weight text, vision and audio models, plus fine-tuning and dedicated deployments for teams with steady traffic.

Not Human Search is a search engine and scoring platform that indexes agent-ready sites and tests whether services, APIs, and tools can be discovered and used by AI agents.

Ultravox is a real-time voice AI infrastructure layer that provides developer-friendly APIs and SDKs to build natural, low-latency voice agents.

OpenRouter routes requests to models from many providers through a single OpenAI-compatible endpoint, with unified billing, automatic failover and side-by-side pricing and latency data.

Jina AI provides search foundation models and APIs, including multilingual and multimodal embeddings, rerankers, and a web reader that converts URLs to Markdown for LLM grounding.

Agentset provides developer infrastructure for building production-ready RAG applications, chat, and search features.

AssemblyAI provides transcription APIs with speaker labels, punctuation and word-level timestamps, plus higher-level models for summarisation, topic detection, sentiment and PII redaction.

Elicit searches a corpus of over one hundred million papers, extracts findings into comparison tables and summarises what a body of research collectively says about a question.

Browserbase runs managed headless browsers so AI agents and scrapers can load real pages, handle sessions and interact with sites without a team maintaining browser fleets.

SkillBoss provides AI agents with access to over 700 APIs and 1,000 skills through a single account and API key.

Groq runs language models on its own LPU hardware, producing token throughput and time-to-first-token figures well beyond typical GPU serving, which matters for voice and interactive agents.

Pinecone stores and searches high-dimensional embeddings at scale, returning nearest matches in milliseconds without the operational work of running a vector index yourself.

LlamaIndex is a framework for retrieval-augmented generation and document workflows, handling ingestion, chunking, indexing and querying over private data.

MinerU is a document parsing platform designed for LLMs, agents, and RAG systems. It handles formats like PDF, Word, PPT, Excel, images, and web URLs, converting them into structured outputs such as Markdown, JSON, and L…