@hippo-core/paperclip-plugin

Persistent memory for Paperclip agents — powered by Hippo Core.

What it does

Every agent in your Paperclip company gets two new tools:

• hippo_recall — retrieves relevant memories at the start of any task
• hippo_remember — stores important decisions, outcomes, and facts

Agents remember past work across sessions. No more re-explaining context.

Install in Paperclip

1. Go to Settings → Plugins → Install Plugin
2. Enter: @hippo-core/paperclip-plugin
3. Click Install
4. Configure your AI provider (works with Ollama, OpenAI, OpenRouter, Groq)

Configuration

| Setting | Default | Description | |---|---|---| | API Key | ollama | Your AI provider API key | | Base URL | http://localhost:11434/v1 | API endpoint | | Chat Model | llama3.2 | Model for memory extraction | | Embedding Model | nomic-embed-text | Do not change after first use | | Memories per recall | 5 | How many memories per task | | Max memory tokens | 500 | Token budget for injected memories |

Works with Ollama (free, local)

Set API Key to ollama and Base URL to http://localhost:11434/v1. Pull the required models:

ollama pull llama3.2
ollama pull nomic-embed-text

Memory storage

All memories are stored locally at ~/.hippo-core/paperclip-memory.db. No data leaves your machine except the LLM API calls you configure.

MIT License