Dimension scores are derived from public data and fields; weighted into the composite. Reference only.
Memory System is a persistent semantic memory layer for AI Agents, designed to let Agents continuously write, read, and retrieve context across sessions, models, and teams. It is not just a vector database; instead, it builds on underlying capabilities such as pgvector to provide typed memories, automatic summaries, tags, permissions, auditing, MCP tools, and a service gateway.
The product emphasizes model independence, with support for Claude, GPT, Gemini, Llama, Mistral, Ollama, as well as Cursor, Cline, and custom Agents. For integration, it offers native Claude Code MCP support and a REST API, allowing Agents to call memory_remember and memory_search. Its “Talks like AI” feature can turn raw notes, Slack threads, or scattered ideas into structured memories such as decisions, patterns, and constraints, making them easier to feed into an LLM context window later. Team features include departmental hierarchy, shared banks across teams, fine-grained CRUD permissions, temporary contractor access, and memory analytics.
The security setup is fairly comprehensive: AES-256-GCM encryption at rest, TLS 1.3, 15-minute JWTs, TOTP MFA, Postgres RLS row-level isolation, per-org keys, audit logs, rate limiting, and privilege escalation approval. By default, it is hosted on AWS us-east-1; Pro and Enterprise plans can choose the EU, while Enterprise supports VPC, on-prem, or air-gapped deployment. The company states that it does not use customer memories to train models and does not share them with model providers unless the user explicitly enables summarization or embedding calls. The Service Gateway allows Agents to call APIs such as Cloudflare and AWS through controlled permissions, without directly holding the underlying secrets.
Starter costs $29/month and is suitable for individual developers. Pro costs $100/month and includes 5 seats, targeting teams with features such as sharing, audit export, webhooks, SSO, and the Service Gateway. Enterprise is custom-priced and covers compliance, dedicated infrastructure, HIPAA BAA, SAML/SCIM, SLA, and dedicated support. All plans include a 14-day free trial with no credit card required.
Its strengths are a clear and focused positioning, straightforward integration options, strong security governance, and the ability to reduce the cost of repeatedly explaining codebases and business decisions to Agents. Limitations include no disclosed Chinese interface or documentation, SOC 2 still being described as a readiness path, and a lack of public benchmarks for real-world retrieval quality and multilingual performance. Access from mainland China and supported payment methods are not disclosed, so china_access can only be considered unknown. If network or compliance constraints apply, alternatives include self-hosting pgvector/Qdrant/Weaviate with MCP, or using options such as LangChain memory.
⚠ This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on kalebdienelt.com official site.
kalebdienelt.com is an Unknown AI Apps provider. TG4G tracks its product information, an overall rating of 7.0/10, and a China-accessibility score of Workable. Click "Visit Official Site" to reach kalebdienelt.com directly.