Use cases

Built for real agents.

Research Agent

POST /v1/search

Ingest 100+ papers, query across them in natural language.

Customer Support

POST /v1/memory

Remember customer preferences and prior issues across sessions.

Multi-Agent Team

POST /v1/shares

Three agents share a knowledge base without sharing API keys.

Capabilities

Everything an agent needs.

01

Semantic file storage

Upload any file and automatically extract, chunk, and embed it for natural language search. PDFs, HTML, JSON, TXT — all become queryable.

pgvector HNSW · 1536-dim · <50ms p99
02

Agent memory

Persistent working memory with TTL support. Store preferences, facts, and context that survives across sessions.

5,000 entries · vector(1536) · TTL support
03

Agent-to-agent sharing

Grant read access to your file embeddings to other agents. Build multi-agent systems that share knowledge, not credentials.

Cross-agent ACL · embedding-level grants
04

Collections & scoping

Group files into named collections. Search within a project, a client, or a time period — never wade through irrelevant results.

Unlimited collections · metadata filtering
05

Event-driven webhooks

Register a URL and get notified when files are uploaded, indexed, or deleted. Build reactive agent workflows.

4 events · SSRF-protected · signed payloads
06

MCP Server

Plug FileVault directly into Claude Desktop, Cursor, Cline, or any MCP-compatible client. No code required.

stdio transport · 5 tools · Zod schemas

Getting started

Four steps to agent storage.

01

Create your agent

One POST request. Get an API key that starts with fv_sk_. It is shown exactly once.

POST /v1/agents
02

Store & index files

Upload via multipart form. We extract text, chunk it, and generate embeddings in the background.

POST /v1/files
03

Search in natural language

Query across files and memory with pgvector-powered semantic search. Results in <50ms.

POST /v1/search
04

Share with other agents

Grant read access to your embeddings. Other agents search your files without seeing your credentials.

POST /v1/shares

Testimonials

We replaced S3 + Pinecone + a custom auth layer with FileVault in a single afternoon. The MCP server means our ops team can query agent memory without writing code.
H

Hemanth A

AI Engineer, Cognizant