Hippo

•
794
•

Provides memory for AI agents, learning from mistakes and replacing old facts with new information.

Hippo tackles a critical flaw in traditional AI memory systems: their inability to learn from negative outcomes. Instead of merely storing and retrieving, Hippo enables AI agents to truly learn what is wrong, what has been replaced, and what is no longer relevant, ensuring that past mistakes are not repeated. It achieves this by allowing users to mark memories as incorrect or superseded, and by strengthening frequently used facts. This system also solves the pervasive problem of memory portability, consolidating diverse agent-specific memories—from Claude Code, Cursor, ChatGPT exports, and Slack history—into a single, persistent SQLite store accessible across any tool supporting MCP or HTTP, thereby providing a unified, outcome-aware knowledge base for all your AI agents.

Key Features

01Extensive Ingestion Capabilities: Imports memories from ChatGPT exports, Claude's CLAUDE.md, Cursor rules, Slack history, markdown, and raw conversation text.
02Outcome-Based Learning: Actively learns from errors and new facts, preventing AI agents from repeating past mistakes.
03Cross-Agent Portability: Unifies memory from various AI agents (e.g., Claude Code, Codex, Cursor) into a single, persistent SQLite store.
04Explainable Recall & Provenance: Provides insights into why a memory was returned and tracks the origin of every stored fact, supporting GDPR compliance.
05Session Management & Continuity: Supports active task snapshots, session event trails, session handoffs, and working memory for seamless workflow continuity.
06764 GitHub stars

Use Cases

01Creating a unified, portable memory system for developers using multiple AI assistant tools across different projects.
02Consolidating diverse knowledge sources—like project documentation, chat logs, and commit history—into a single, intelligent repository for AI agents.
03Enhancing AI coding agents with adaptive memory that prevents repeating past errors and improves decision-making.