PLUR MCP - Persistent Memory for AI Agents

PLUR solves the amnesia problem: AI agents start every session with amnesia - you re-explain the project, repeat your preferences, and correct the same mistakes over and over. PLUR gives them a memory that persists.

It's open, local-first memory for AI agents. Your agent's corrections, preferences, and conventions are stored as plain-text engrams on your own machine - memory you can read, correct, and carry across sessions without vendor lock-in.

What It Does

• Persistent agent memory: Corrections, preferences, and conventions survive across sessions
• Local-first: All engrams stored as plain-text files on your machine - no cloud dependency
• Open source (MIT): Full transparency, no vendor lock-in
• Multi-agent aware: One memory store for all your MCP-compatible agents
• Human-readable: Engrams are plain text - you can read, edit, and prune them directly

Key Tools

Tool	Description
plur_store	Save a new memory engram
plur_search	Semantic search across stored engrams
plur_list	List all stored engrams
plur_delete	Remove an engram by ID

Quick Start

1. Install

git clone https://github.com/plur-ai/plur
cd plur
npm install
npm run build

2. Configure Your MCP Client

{
  "mcpServers": {
    "plur": {
      "command": "node",
      "args": ["/path/to/plur/dist/index.js"],
      "env": {
        "PLUR_STORE_PATH": "~/.plur/engrams"
      }
    }
  }
}

3. First Use

Once connected, your agent can call plur_store to remember anything: - "Remember I prefer TypeScript with strict mode" - "Store that our API base URL is https://api.corpusiq.io" - "Save the convention: always use async/await, never .then()"

On the next session, call plur_search and the agent retrieves what it needs.

Hermes Agent Integration

# ~/.hermes/config.yaml
mcp_servers:
  plur:
    command: "node"
    args: ["/home/hermes/tools/plur/dist/index.js"]
    env:
      PLUR_STORE_PATH: "~/.plur/engrams"

Use Cases for Business Operators

1. Self-improving agents: Your growth agent remembers what content performed well, what outreach templates worked, and which communities respond best
2. Cross-session consistency: Agents carry preferences, style guides, and business rules across restarts
3. Team memory: Share engram stores across team members for consistent AI behavior
4. Audit trail: Plain-text engrams let you inspect exactly what your agents have "learned"

How It Compares

Tool	Approach	Storage	License
PLUR	Local-first engrams	Your filesystem	MIT
Groundwork	Company-state feed	Cloud-hosted	Source-available
Memory (official)	Knowledge graph	Local file	MIT
Context7	Doc retrieval	Cloud	Proprietary

PLUR differentiates by being: (1) fully local, (2) plain-text readable, (3) MIT licensed, and (4) purpose-built for agent corrections and preferences rather than general knowledge retrieval.

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations

• No cloud sync (yet): Engrams are local - share via git, rsync, or your own sync solution
• New project: Early-stage, API may evolve
• Text-only: No structured data or multimodal memory (yet)
• Local setup required: Must install Node.js and clone the repo - no hosted version

Category

Memory & Knowledge - PLUR focuses specifically on agent corrections and preferences, making it complementary to general documentation tools like Context7 and company-state feeds like Groundwork.

---

Discovered via mcp.so on July 24, 2026. Verified listing. GitHub: plur-ai/plur. 226 stars.