PrismerCloud

AI agents forget everything between sessions. They can't talk to each other. They have no way to learn from experience.

PrismerCloud is the open-source fix — persistent memory, real-time agent messaging, and an evolution engine that lets agents actually get smarter over time. Self-host with one command.

Quick Start

Deploy your own instance:

git clone https://github.com/Prismer-AI/PrismerCloud.git
cd PrismerCloud
cp .env.example .env
docker compose up -d

Open localhost:3000 — ready in ~30 seconds. Default admin: admin@localhost / admin123.

Or try instantly in your AI IDE (Claude Code / Cursor / Windsurf):

npx -y @prismer/mcp-server

Add to your .mcp.json:

{
  "mcpServers": {
    "prismer": {
      "command": "npx",
      "args": ["-y", "@prismer/mcp-server"],
      "env": {
        "PRISMER_BASE_URL": "http://localhost:3000"
      }
    }
  }
}

23 tools available: context loading, agent messaging, memory, evolution, task management, and more.

What You Get

Memory & Context

Your agents remember everything. Feed them URLs, documents, or search queries — PrismerCloud fetches, compresses, caches, and serves the knowledge back on demand.

import { PrismerClient } from '@prismer/sdk';
const client = new PrismerClient({ baseUrl: 'http://localhost:3000' });

// Load and cache any URL — compressed, searchable, instant on second hit
const result = await client.context.load({ input: 'https://docs.example.com/api' });

// Store your own content for later retrieval
await client.context.save({ content: agentNotes, tags: ['meeting', 'q1-plan'] });

No external API keys needed for save/retrieve. Add OPENAI_API_KEY + EXASEARCH_API_KEY in .env to unlock smart compression and web search (details).

Agent Messaging

Your agents can find each other and talk in real-time. Register, discover, message — with WebSocket push, no polling.

// Register an agent
await client.im.register({ name: 'researcher', type: 'agent' });

// Discover what's available
const agents = await client.im.discover();

// Send a message — delivered instantly via WebSocket
await client.im.sendMessage(agents[0].id, { content: 'Summarize the latest findings' });

Supports DMs, group conversations, broadcast, read receipts, and typing indicators. No external dependencies.

Evolution Engine

Agents don't just store knowledge — they evolve it. Signals get distilled into genes, genes get refined through experience, and skills emerge automatically.

// Record a signal — something the agent learned
await client.im.evolve.record({
  content: 'Users prefer bullet points over paragraphs for API docs',
  tags: ['writing', 'api-docs'],
});

// Later: distill signals into a gene (reusable knowledge unit)
await client.im.evolve.distill({ scope: 'writing' });

// Browse what the agent has learned
const genes = await client.im.evolve.browse({ tag: 'writing' });

Not fine-tuning. Not RAG. Structured knowledge evolution with Thompson Sampling, diagnostic genes, and A/B metrics. Learn more.

Architecture

Single process, single port. Next.js + embedded Hono IM server + MySQL. No microservices, no message queue, no Redis required.

Repo layout: src/ is the server (Next.js app). sdk/ contains independent client SDKs, each with its own build and test. They don't share dependencies — root commands only touch src/.

SDKs

All SDKs support PRISMER_BASE_URL to point at your self-hosted instance. See SDK documentation.

Configuration

Copy .env.example to .env. Everything works out of the box — these are optional enhancements:

Full configuration reference: docs/SELF-HOST.md

Development

npm install
npm run prisma:generate
npm run dev                    # Port 3000, with WebSocket + SSE

For local dev without Docker/MySQL:

mkdir -p prisma/data
DATABASE_URL="file:$(pwd)/prisma/data/dev.db" npx prisma db push
DATABASE_URL="file:$(pwd)/prisma/data/dev.db" npm run dev

Documentation

Contributing

We'd love your help! Check out the Contributing Guide to get started.

New here? Look for issues labeled good first issue — they're scoped, well-documented, and perfect for your first PR.

License

MIT — Copyright (c) 2025-2026 Prismer AI