Shared memory for LlamaIndex
LlamaIndex is the go-to framework for connecting LLMs to your data, indexing documents, building retrieval pipelines, and powering agents that reason over a knowledge base. It is excellent at retrieving from static corpora you have ingested, but the agents you build on it still forget the dynamic, run-to-run facts they generate, what a user prefers, what a previous agent concluded, what changed since the last index build. That working knowledge is separate from your document index, and LlamaIndex leaves you to persist and share it yourself. Glen, shared memory for AI agents, provides exactly that layer as a single MCP tool your LlamaIndex agents read from and write to.
Give your LlamaIndex agents access to Glen over MCP and each agent gains one tool that retrieves accumulated context and records new facts in a single round trip. Your retrieval pipeline keeps doing what it is good at, pulling relevant passages from your indexed documents, while Glen holds the evolving, agent-generated knowledge that does not belong in a static index: decisions, preferences, observations, the things learned during a run. Before an agent acts, it can read what the organization already knows; after it acts, it writes back what it discovered, with no separate persistence service to build.
The distinction matters: a LlamaIndex index answers what the documents say, and Glen remembers what the agents have learned. Because Glen is org-scoped, that learned memory is shared across every agent and pipeline in your organization rather than trapped in one process, so an agent built today inherits what an agent built last month recorded. And because Glen is a standard MCP server, the same memory is readable from Claude Code, Cursor, or any other MCP client, so your retrieval agents and your developers draw on one shared store. Connect once over OAuth or an API key and let the agent-side knowledge compound alongside your document retrieval.
FAQ
- Does Glen replace my LlamaIndex vector store?
- No. Your index retrieves from documents you ingested. Glen holds dynamic, agent-generated knowledge, decisions, preferences, observations, that does not belong in a static index. They complement each other.
- How do LlamaIndex agents connect to Glen?
- Glen is a standard MCP server, so your agents call it as an MCP tool to read shared context and write new facts, authenticated over OAuth or an API key.