Memanto
AI agent memory software preserves context across Claude Code, Cursor and Codex, with local Docker and Ollama operation under the MIT license.

Memanto is a companion memory agent for developers whose AI assistants lose project decisions and preferences between sessions. It manages shared knowledge across tools such as Claude Code, Cursor and Codex, then briefs agents when they start work.
- Memory curation: Extract durable facts from conversations, merge duplicates and reconcile conflicting information while retaining the history of earlier beliefs.
- Relevant recall: Retrieve context for the current task, favor fresh information and answer questions using stored memories.
- Scoped access and sources: Give agents separate namespaces and trace memories through their sources, timestamps and confidence scores.
- Portable knowledge: Inspect memories in a local dashboard or export readable Markdown in Open Knowledge Format; migration works with Mem0, Letta, Supermemory, Zep and Hindsight.
- Retention and history: Policies can mark memories as expired without deleting their contents, and point-in-time recall can reconstruct earlier beliefs; expired entries remain labelled in recall and can be restored.
Memanto is free and open source under the MIT license and runs on macOS, Linux and Windows. Its local mode uses Docker and Ollama for embeddings and answers, with no account, API key or outbound calls from the memory system. Moorcheh Cloud is an optional managed backend that requires an API key and moves backend processing to a cloud service. OpenAI and Cohere are also optional external providers.