About Antfly
Antfly is a retrieval engine that grounds AI apps and agents in an organization's own knowledge. It combines vector, keyword and graph search as indexes inside one database, fused in a single query plan, and performs chunking, embedding, indexing and reranking inside the engine as data arrives, so there is no separate pipeline to assemble or tune.
It is built for developers who want retrieval for agents without wiring together a vector store, an embedding service and a reranker. Antfly sits on top of existing Postgres databases, storage buckets, file systems and document stores without migration, and handles PDFs, screenshots, tables and audio. Embedding, reranking and entity extraction models ship inside the engine, including multimodal embedders for text with images and for text with audio and a zero shot recognizer for entities and relations, and none of that inference leaves the customer's infrastructure. A single node engine installs locally with one command. Projects built on it include SearchAF, a free macOS app for on device semantic search across personal files.
The open source edition is free with the full feature set, hybrid BM25 and vector search, built in inference, multimodal support and no limits on data or queries. Growth is an invite only cloud plan starting at $79 per month per private instance: Starter with 1 CPU, 4 GB of memory and 50 GB of storage at $79, Standard with 2 CPU, 4 GB and 100 GB at $149, and Pro with 4 CPU, 8 GB and 200 GB at $299, with a high availability option. It adds RBAC, SSO, accelerators, autoscaling and managed backups, and inference credits are prepaid at $0.01 each. Queries are never metered. Enterprise adds a custom SLA, volume discounts and SOC 2 and HIPAA compliance.