# Memanto [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/moorcheh-ai/memanto  
**GitHub Stars:** 2321  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/memanto

## Description
MCP server for Memanto - persistent semantic memory for any MCP-compatible agent

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "memanto": {
    "command": "uvx",
    "args": ["memanto"]
  }
}
```

## Documentation

## What Memanto does

The Memanto MCP server provides managed semantic memory for MCP-compatible agents. It is designed to sit beside an agent fleet as a memory-focused companion, rather than acting as a passive transcript archive. Memanto extracts durable knowledge such as decisions, preferences, facts, and failures from agent conversations, then maintains that knowledge over time.

Its memory estate can be shared across agents so that one agent can recall information learned by another. The system distinguishes current knowledge from earlier beliefs, keeps provenance and timestamps, and supports confidence tracking. Agent namespaces provide separation between different agents, such as production operations and experimentation.

Memanto belongs in the Knowledge & Memory category for workflows where agents need continuity across sessions, users, or tools. It is especially relevant when duplicated memories, contradictory information, stale preferences, or repeated work make an unmanaged memory store difficult to use.

## How it works

Memanto observes interaction streams and extracts selected durable knowledge instead of retaining every conversation as memory. Scheduled processing consolidates duplicates, joins related fragments, and increases confidence when observations repeat. When new information conflicts with an older belief, the newer knowledge can supersede it while preserving the historical record.

Memory lifecycle policies control decay and expiry. Expired memories remain available for recall with an expiration marker and can be restored; explicit deletion is separate. The `memanto conflicts` command supports reviewing contradictions, while `memanto forget` handles forgetting operations.

Before an agent acts, `memanto agent bootstrap` supplies a relevant slice of the estate. Agents can also use commands such as `memanto recall` for searches and `memanto answer` for grounded responses. Queries can be evaluated as of a date or against changes since a named release, where those options are needed for historical review.

## Setup and configuration

Install the package with Python, npm, or the Go SDK:

- `pip install memanto`
- `npm install @moorcheh-ai/memanto`
- `go get github.com/moorcheh-ai/memanto/sdks/go`

Running `memanto` starts the setup path for either an on-premises deployment or the cloud option. The on-premises mode uses Docker and Ollama and is described as requiring no account or API key. `memanto config backend` switches between configured backends.

Connect supported agents with `memanto connect`, followed by a client name such as `claude-code`, `cursor`, `codex`, `windsurf`, or `cline`. The repository also documents connections for Goose and Copilot, although those are not represented in the directory's supported-client list. `memanto ui` opens a local dashboard for browsing, searching, and auditing the estate.

## Tools and capabilities

The Memanto MCP server includes capabilities for:

- Remembering knowledge from text or conversations with `memanto remember`.
- Recalling relevant information with ordinary queries, date-based views, or change-based views.
- Producing grounded answers through `memanto answer`.
- Consolidating memories through scheduled processing with `memanto schedule enable`.
- Reviewing contradictions with `memanto conflicts`.
- Applying configurable retention and expiry policies.
- Exporting memory as Markdown through the Open Knowledge Format with `memanto memory export --okf`.
- Importing or migrating memory from Mem0, Letta, Supermemory, Zep, Hindsight, or OKF bundles.

## Limitations and notes

The README describes on-premises operation with Docker and Ollama, plus a cloud deployment, but does not specify detailed hardware, model, or hosting requirements. Client connection support is shown through the Memanto CLI rather than a complete transport or configuration reference in the supplied material.

The Memanto MCP server manages memory but does not eliminate the need to define retention policies and review conflicts. Expiry is policy-driven, and expired records remain visible in recall until deleted or restored. Open Knowledge Format export is intended to keep the estate portable as readable, diffable Markdown rather than a proprietary archive.

Memanto is released under the MIT License.

_Full upstream README: https://allmcps.com/mcp/memanto/readme_