Memory that survives the session
Store task state, model notes, tool outputs, files, and retrieval history so agents can pick work back up cleanly.
Private AI memory for product teams
Give your AI agents a memory customers can trust.
StateVault helps agent products remember work across sessions, keep customer data separated, show clear usage evidence, and export memory when customers need to leave.
- Customer-owned
- Separate vaults for each account
- Private by design
- Zero-training data protection
- Easy to leave
- Portable ZIP exports
Private-beta proof
Live API and MCP access, customer-scoped vaults, portable exports, and usage evidence are already working.
The need
AI products need memory that feels trustworthy, not improvised.
Modern agents need more than prompts and chat history. They need a dependable place to remember work, find the right context, stay inside the right customer boundary, and prove what happened.
StateVault gives teams that layer without tying customer memory to one model vendor. Use the API or MCP, keep accounts separated, review usage, and provide exports from the start.
Who it serves
Built for teams turning agents into products customers rely on.
Shared memory without messy boundaries
Coordinate agent workflows while keeping each customer, project, and vault clearly separated.
A safer home for proprietary workflows
Keep sensitive agent memory outside model-training loops with exportable state and conservative private-beta controls.
How it works
Customer vaults with search, files, usage evidence, and exports.
01
Separate customer vaults
Each customer account can keep memory in scoped vaults, with stronger isolation reviewed by an operator when needed.
02
Search that respects boundaries
Context search stays inside the right customer and vault so agents can find useful memory without widening exposure.
03
Exports from the beginning
Files and export packages stay tied to the right customer and vault, giving buyers a practical path out if they need one.
Interoperability
MCP turns the vault into a universal agent interface.
Instead of every developer wiring memory by hand, StateVault exposes it through API and MCP patterns agent systems can already use.
Resources
Read recent memory, files, records, and context bundles.
Tools
Write memories, update state, request exports, and manage lifecycle events.
Prompts
Bring relevant context back into the current workflow without manual prompt stitching.
Trust layer
Designed for buyers who treat AI memory as customer data.
Zero-training guarantee
Customer state is not used to train foundation models. The value is custody and continuity, not data extraction.
Customer separation
Authentication, roles, customer accounts, vaults, and reviewed isolation keep memory surfaces separated.
Private-beta controls
Role boundaries, login-first portals, maintenance controls, and operator-reviewed provisioning keep sensitive workflows gated.
Portable export
A dedicated export path lets customers retrieve their memory history and reduces lock-in risk.
Private beta
Core memory workflows are ready for controlled private-beta evaluation.
StateVault now has the working loop behind the promise: API state operations, MCP access, scoped API keys, context search, customer/vault usage evidence, portable ZIP exports, and live isolation proof.
REST API
MCP server
Scoped API keys
Isolated vault runtime proof
Customer/vault usage evidence
Semantic retrieval
Portable ZIP exports
Paid pilot model
Simple usage evidence before live payment collection.
Private-beta usage is measured by customer, vault, method, operation, and billing period. Live payment collection remains disabled until a separate activation milestone.
Entry
Durable memory, basic context retrieval, and a default customer vault.
Pro
Expanded vault usage, stronger context search, portable exports, and operator-reviewed isolation.
Enterprise
Dedicated-vault operating review, custom retention discussions, and priority support planning.
Plain answer
What to know about StateVault.
What is StateVault?
StateVault is a private memory layer for AI agents. It stores state, files, metadata, and context so agent products can continue work across sessions, models, tools, customers, and vaults.
What makes it different from generic cloud storage?
The product is organized around customer vaults, agent memory operations, context search, MCP interoperability, usage evidence, and exportable memory history.
Why independent infrastructure?
Independent memory helps reduce model-vendor dependency and gives sensitive AI workflows a clearer customer-control story.
Early access
Build agent memory your customers can trust.
Private beta developer docs are available by request for teams evaluating customer-scoped agent memory, many vaults per account, MCP interoperability, context search, usage evidence, and portable exports.