Skip to main content
Pixel brand continuity across app, CLI, and docs

The same robot that guides vai now carries the docs.

The fastest path from documents to semantic search. The pixel robot already gives the CLI a memorable set of moments for greeting, thinking, searching, and success. Bringing those same poses into the docs makes the whole brand feel like one product family instead of separate surfaces.

OnboardingGuidesRelease notes
npm install -g voyageai-cli
Get Started5-Minute RAG PipelineGitHub
robot/wave
vai pixel robot waving

A shared visual language for hello, in-progress work, and shipped results.

wavethinkingsuccess

Brand imagery

Carry the same moments from the terminal into the docs

The CLI already gives the robot a job: greet people, show work in flight, and celebrate clean finishes. The docs can reuse those same visual cues in the places users learn, search, and ship.

wave

Wave people into the product

The same wave pose used for CLI greetings gives docs entry points the same friendly first-run energy.

Start with installation

thinking

Show thinking where work is in flight

Guides, explain pages, and walkthroughs can reuse the search and thinking language users already see in the CLI.

Explore explain topics

success

Land on success moments

Release notes, quickstarts, and finish states feel more cohesive when the docs celebrate successful outcomes the same way the tool does.

Read the latest release

CLI: 33 Commands

Embed, search, rerank, ingest, evaluate, benchmark, and more. One tool for the entire RAG pipeline, from chunking documents to querying with reranking.

MCP Server: 11 Tools

Drop vai into Claude Desktop, Cursor, or any MCP-compatible editor. Your AI assistant gets semantic search, embeddings, reranking, and cost estimation.

Agentic Workflows

Define multi-step RAG pipelines as portable JSON files. Search multiple collections, merge results, filter by relevance, and summarize with LLMs.

From zero to semantic search in 5 minutes

$ vai init
Created .vai.json with defaults

$ vai pipeline ./docs --query "how does auth work?"
Chunking 12 files... 47 chunks
Embedding with voyage-3-large... done [1.2s]
Storing in MongoDB Atlas... 47 documents
Searching... 10 candidates
Reranking with rerank-2.5... done [0.3s]

Top results:
[1] docs/auth/overview.md (0.97)
[2] docs/auth/jwt-tokens.md (0.94)
[3] docs/api/middleware.md (0.89)

Everything you need for RAG

Voyage AI Embeddings

State-of-the-art embedding models including voyage-3-large, domain-specific models for code, finance, law, and multilingual content.

MongoDB Atlas Vector Search

Store embeddings and run vector searches directly against your Atlas cluster. Automatic index management included.

Two-Stage Retrieval

Combine fast vector search with Voyage AI reranking for dramatically better result quality.

Chat with Your Docs

RAG-powered chat using Anthropic, OpenAI, or local Ollama models. Conversation history and agent mode included.

Evaluation and Benchmarks

Test retrieval quality with custom test sets, compare configurations, and benchmark embedding performance.

30 Educational Topics

Built-in explanations of embeddings, vector search, RAG, MoE architecture, quantization, and more.

vai is a community project by Michael Lynn. It is not an official MongoDB or Voyage AI product. Use of Voyage AI APIs and MongoDB Atlas requires accounts with those services.