An innovative self-trained conflict judge
mema · an AI assistant with a clean memory mema-twin · a self-evolving, self-learning AI assistant

Give every Agent lasting memory

Models set the ceiling; memory sets the slope.

How do you raise the slope? Execution · memory · collaboration

A long-term memory layer for AI Agents: project facts and preferences survive every session, conflicts are judged on the spot, recall lands on the exact sentence.

Ask your Agent to installExplore Team editionEnter Team Portal
How to raise the slope

More than memory: a big lift for your Agent

mema bolts three layers onto any Agent — execution, memory, collaboration. Execution makes it work the way you work; memory makes it remember you; collaboration gives every Agent one shared, trusted memory. Plug in once, get stronger every day.

mema-twin · your twin

Execute

More than experience: discipline from day 0

  • Task decomposition: dependency-aware step plans; anti-skip gates stop headless sprints
  • Loop discipline: every failure logs a reflection; retries and fallbacks archived for reuse
  • Tool pitfalls: unusual paths go to tool_log, distilled into playbooks injected next time
mema · all the infrastructure

Memory

Personal memory — and a knowledge base

  • Carries project context, preferences, and decisions across sessions
  • Docs, conclusions, pitfalls — store them all, use it as a knowledge base
  • Hybrid vector + full-text retrieval, sentence-precise recall
mema · the shared memory base

Collaborate

Many Agents, one trusted memory

  • Any MCP client connects with an identity, sharing one library
  • Memory follows you, not the vendor’s Agent
  • Conflicts judged on the spot; governance keeps memory clean

How big is the lift: the compounding curve

Performance Sessions / days 95 Top-tier Agent · no memory Cold start · discipline included Ordinary Agent + mema · accumulated, above from the start Cold start overtakes eventually too
Ordinary Agent + mema: accumulation carries over — starts above the top and keeps compounding Cold start (empty library): decomposition, retry, and verification discipline from day one Top-tier Agent without memory: climbs back to 95 from zero every session — a flat line in expectation
Read the full argument →

The amnesia problem · the four-layer stack · the compounding curve · honest limits

mini-clash · self-trained judging core

Write-time conflicts: the self-trained judge decides on the spot

Write quality · recall precision

The write gate and sentence-level recall

mema ecosystem

mema-twin: your work self, compiled into a prompt

Built on mema Apache-2.0

mema-twin

Your work self, compiled into a prompt — every Agent starts the day as you.

Templates, metrics, tone, taboos — collected one by one, compiled overnight into a versioned persona.

plan-mode

Apache-2.0

agent-lane

Apache-2.0
Get started

One prompt. Let your Agent install it.

Copy the task into Codex, Claude Code, Cursor, or any Agent with terminal access.

Recommended

Copy the task and paste it into your Agent

It will detect your OS, Python environment, and current MCP client, then choose the transport for your setup: stdio for one Agent or HTTP for multiple Agents. Existing config and databases stay intact, and installation ends with a doctor check.

  1. 01Copy the install task
  2. 02Paste it into your Agent
  3. 03Review the doctor result
Install task for your Agent 
Read the latest instructions at https://github.com/billy12151/memory-arbiter-mcp#install-with-your-ai-agent. Install and configure the latest mema release for my operating system and current AI client.
Confirm how many Agents I need to connect and choose the appropriate MCP transport: prefer local stdio for a single Agent and an HTTP MCP server for multiple Agents.
Preserve any existing config and database; do not overwrite or delete existing data. Ask me before choosing between materially different install modes, changing existing config, or performing any destructive or privileged action.
When finished, run mema doctor and report the install method, config path, database path, client integration, and verification result.

The Agent uses the GitHub README as its source of truth and asks before changing existing config, choosing a material install mode, or performing a sensitive action.

Manual install and every option live on the Overview page →

One-page overview

The key content beyond the animations

Six cards, six chapters of the full introduction — each links straight there.

01

Core benefits

Three things change once mema is in place: results, tokens, handoff cost.

 02

Editions compared

Personal and Team: one memory core, different deployment and governance boundaries.

 03

Why not plain RAG

Not just storing more records — memory that is usable, trustworthy, and governable.

 04

Onboarding paths

Personal installs locally, Team connects with a Key — any MCP client works.

 05

Honest limits

Where it can't help: no model capability, no tool calls, no context management.

 06

Security boundary

Sharing is an authorization chain: owner + grant + team audit.