Stash (GitHub Repo)

AI agentsinfrastructure Read original

Stash is an open-source tool that gives AI agents persistent memory across sessions, solving the problem of LLMs starting every conversation from scratch.

What: Stash is a self-hosted memory layer for AI agents that uses an 8-stage consolidation pipeline to transform raw observations into structured knowledge including facts, relationships, causal links, goal tracking, and failure patterns, compatible with any MCP-enabled agent.

Why it matters: Addresses a fundamental limitation of LLMs by enabling agents to build up knowledge over time rather than resetting with each interaction, potentially making them progressively more useful as they accumulate experience and learn from past sessions.

Takeaway: Clone the repo and run docker compose up to add persistent memory to any MCP-compatible AI agent like Claude Desktop, Cursor, or Cline.

Deep dive

Uses Postgres with pgvector as the underlying storage engine for vector-based memory retrieval
Implements an 8-stage consolidation pipeline that progressively refines raw observations: episodes → facts → relationships → patterns → wisdom
Each consolidation stage only processes new data since the last run, making it efficient for continuous operation
Includes advanced features like causal link analysis, goal tracking, failure pattern recognition, hypothesis verification, and confidence decay over time
Runs as an MCP server with background consolidation, meaning the memory processing happens automatically without blocking agent interactions
Works with a broad range of AI platforms including Claude Desktop, Cursor, Windsurf, Cline, Continue, OpenAI Agents, Ollama, and OpenRouter
Self-hosted design means you maintain control over your agent's memory data rather than relying on third-party services
Apache 2.0 licensed, allowing commercial use and modification

Decoder

MCP: Model Context Protocol, a standard interface that allows AI agents to connect to external tools and data sources
pgvector: PostgreSQL extension for storing and querying vector embeddings, enabling semantic similarity search
Consolidation pipeline: Multi-stage process that transforms raw data into increasingly abstract and useful knowledge structures

Original article

Stash

Your AI has amnesia. We fixed it.

Every LLM starts every conversation from zero. Stash gives your agent persistent memory — it remembers, recalls, consolidates, and learns across sessions. No more explaining yourself from scratch.

Open source. Self-hosted. Works with any MCP-compatible agent.

Quick Start

git clone https://github.com/alash3al/stash.git
cd stash
cp .env.example .env   # edit with your API key + model
docker compose up

That's it. Postgres + pgvector, migrations, MCP server with background consolidation — all in one command.

What It Does

Stash is a cognitive layer between your AI agent and the world. Episodes become facts. Facts become relationships. Relationships become patterns. Patterns become wisdom.

An 8-stage consolidation pipeline turns raw observations into structured knowledge — facts, relationships, causal links, goal tracking, failure patterns, hypothesis verification, and confidence decay. Each stage only processes new data since the last run.

Works with Claude Desktop, Cursor, Windsurf, Cline, Continue, OpenAI Agents, Ollama, OpenRouter — anything MCP.

Learn More

alash3al.github.io/stash →

License

Apache 2.0