Memory for
your AI agents.
Give any agent persistent, entity-scoped recall in three lines of Python, TypeScript, Go, or .NET. Aether remembers facts, recalls the right context, and includes the database underneath when you need source-grounded retrieval.
Your agent forgets the user between sessions. This one doesn't.
Sarah tells her agent one thing. A new session later — chat history wiped — it still knows. That's the Memory SDK: remember once, recall anywhere. Two calls, first-party in TypeScript, Python, Go, and .NET.
// waiting for the agent…
Agent Economics
Agent memory with one predictable bill.
Aether keeps the Memory API at the center: remember facts, recall context, and retrieve source-backed answers while the vector index, document storage, embeddings, and reranking are bundled behind the platform.
Monthly Cost Behind Agent Memory
Pro-capacity workload: 1,000,000 vectors, 15GB storage, 100,000 recalls/month.
Unified Memory API
One SDK surface for remember, recall, source-backed retrieval, and tenant-scoped context across a fleet.
Bundled Storage & Retrieval
Documents, vectors, embeddings, and reranking are priced as one platform instead of separate services to meter and tune.
Predictable Usage Controls
Plan capacity and overages are explicit, so launch traffic can be modeled before it surprises your margin.
Built for Managed Agent Fleets
When you deploy a fleet of managed agents, memory can become a hidden stack: vector reads, object storage, embeddings, and reranking. Aether keeps those pieces behind the Memory API and prices them as one platform, so teams can model fleet growth before launch.
- Plan-based memory pricing — no separate vector or embedding vendor bill
- Shared memory across your entire agent fleet
- Tenant-isolated memory with auditable source references
What Your Agents Get
Memory is more than a vector call. Aether gives your agents identity, source storage, and retrieval behind the same API.
Control Plane
Access & Identity
Give every agent its own scoped API key and keep tenant boundaries explicit. Your team can trace which agent accessed which source and when, without building a separate control plane.
Document Layer
Tamper-Proof Storage
Source documents are encrypted at rest and tied to content hashes, so retrieval can point back to the material your agents used instead of orphaned vector chunks.
Vector Engine
Low-Latency Retrieval
High-performance vector indexes keep recall fast enough for interactive agents, while source-grounded retrieval keeps the answer path understandable.
Use Cases
What Builders Are Shipping
Managed Agent Memory
Deploy a fleet of managed agents that can share scoped context without copying prompts, vectors, and documents between tools. Works with Claude, GPT, Gemini, or any LLM provider.
Support Agent Context
Remember customer preferences, product facts, and open issues so support agents can pick up useful context across sessions without relying on chat history alone.
Source-Backed Knowledge Bases
Store source documents alongside memory so answers can point back to the material that shaped them, instead of returning detached vector matches.
Auditable Retrieval
When agents make decisions, teams need a source trail. Aether helps connect recalled context to tenant-scoped documents and access history.
Security Without Compromise
Enterprise AI demands more than an API key and a prayer. Aether is built with defense-in-depth — from the hardware layer up.
Verified Compute
Every vector operation is strictly validated for accuracy and performance. You never have to trust that the infrastructure did the math correctly — Aether guarantees deterministic results for every query.
Immutable by Default
Documents stored in Aether cannot be silently modified or deleted. When data is removed from active indexes, the original record persists in cold storage — creating a complete, auditable history of every change your AI knowledge base has ever seen.
Tenant-Isolated Encryption
Your data is encrypted before storage and isolated per tenant through indexing and retrieval. Scoped keys, encrypted records, and auditable access paths keep memory boundaries explicit without forcing your team to run separate infrastructure.
Scoped key
Encrypted memory
Tenant index
Ready Before Your Agents Are
No cold-start problem. Aether is production infrastructure — live, distributed, and ready for your first agent's first API call. Start on the free tier. Scale when your fleet grows.
Production Cloud Footing
Aether runs as a managed API with automated deploys, tenant isolation, and operational monitoring. For launch, the trust story is simple: use the SDK, get API keys, and let Aether operate the memory layer.
Global Retrieval Layer
Source documents are content-addressed, indexed for retrieval, and served through the managed platform. Your team gets redundancy and operational controls without configuring regions, replicas, or separate storage pipelines.