moorcheh-ai/memanto

Memory that AI Agents Love!

What it solves

Memanto is a Memory Agent designed to manage the long-term memory of a fleet of AI agents. While standard vector stores only persist data, Memanto actively curates it—resolving contradictions, merging duplicates, and expiring stale information—to prevent the "noise" that typically accumulates in agentic memory over time.

How it works

Instead of acting as a simple API, Memanto runs as a companion agent that observes agent interaction streams to extract durable knowledge. It uses a typed semantic memory system (categorizing memories into types like decision, preference, or fact) and an information-theoretic retrieval engine. It manages the memory estate through a set of automated behaviors:

  • Extraction & Consolidation: Pulls knowledge from conversations and merges fragments into a canonical estate.
  • Reconciliation: Detects and versions contradictions when new knowledge conflicts with old.
  • Managed Forgetting: Executes expiry policies (e.g., retiring a context memory after 7 days) to keep recall sharp.
  • Briefing: Provides agents with a minimal, relevant slice of memory before they act, rather than requiring agents to query a database.
  • Portability: Uses the Open Knowledge Format (OKF), which stores the estate as portable, human-readable Markdown files.

Who it’s for

Developers building fleets of AI agents across different frameworks and vendors (such as Claude Code, Cursor, and custom stacks) who need a shared, auditable, and portable memory layer that doesn't lock them into a specific platform.

Highlights

  • Cross-Platform Sync: Connects to various tools like Claude Code, Cursor, Windsurf, and GitHub Copilot via a single command.
  • Local-First Sovereignty: Can run entirely on-prem via Docker and Ollama with no outbound calls.
  • Temporal Recall: Supports point-in-time queries to see what the fleet believed at a specific date or since a specific version.
  • Policy-Based Expiry: Customizable YAML policies to pin important memories or automatically expire low-confidence guesses.
  • Open Format: Memory is stored in OKF (Markdown), making it diffable, greppable, and easy to migrate between providers.

Related

  • Project
  • Project
  • Project
  • Project
  • Project