AgentMem

AgentMem is a memory layer for AI agents. It is built for developers who want agents to keep context instead of starting from zero, and it describes its purpose as giving persistent, searchable memory that can sync across frameworks including LangChain, CrewAI, OpenAI, Claude, AutoGPT, and LlamaIndex.

The product stores and retrieves memories through an API. Its listed capabilities include semantic search, which finds memories by meaning rather than keywords and is powered by vector embeddings, and cross-agent sync for sharing context between agents. The page also says memories are persistent and survive restarts, that the API is simple to use with an API key, and that memories are auto-embedded when stored. A short code example shows calls for storing a memory, searching memories, and syncing across multiple agents.

AgentMem’s production-oriented features include fast API response times under 100 ms, edge caching, API key authentication, tenant isolation, encrypted-at-rest storage, usage analytics for memory usage, search patterns, and agent activity, and a global CDN setup. It is deployed on Railway with a Cloudflare proxy. The page also includes a worked example of registering an agent through a POST endpoint and then using POST endpoints to store memories and search them.

Pricing is presented in three plans. A free plan costs $0 per month and includes 10,000 memories, semantic search, support for three agents, and community support. Pro costs $29 per month and adds unlimited memories, unlimited agents, cross-agent sync, priority support, and usage analytics. Enterprise costs $199 per month and adds a self-hosted option, custom integrations, an SLA guarantee, and dedicated support. The page identifies AgentMem as being built by Gritza and says it is in public beta.