Doesn't the Agents SDK already have sessions?

Sessions persist conversation state for a given run. Glen is durable, org-shared long-term memory, read and written by every agent and MCP client in your organization, that outlives any single run or session.

How do I connect Glen to an Agents SDK agent?

Glen is a standard MCP server, so you attach it as an MCP tool the same way you would any other, authenticating over OAuth or an API key.

Shared memory for the OpenAI Agents SDK

The OpenAI Agents SDK gives you a clean way to build agents with tools, handoffs, and guardrails, and it speaks MCP, so you can attach MCP servers as tools an agent can call. What it does not give you is durable, shared memory: each agent run starts fresh, sessions hold conversation state but not cross-run knowledge, and anything one agent learns is lost to the next unless you build the persistence yourself. Glen, shared memory for AI agents, fills that gap as a single MCP tool your Agents SDK agents read from and write to, backed by one organization-wide store.

Attach Glen as an MCP server in the OpenAI Agents SDK and your agent gains one tool that both retrieves relevant long-term context and records new facts in a single round trip. Before an agent decides or hands off, it can pull what your organization already knows about the task; as it completes work, it writes durable observations back. You stop standing up a vector database, an embedding pipeline, and a bespoke read/write path just to give your agents memory that survives the run, and you stop isolating each agent's learning.

Because Glen is org-scoped, that memory is shared across every agent you build on the SDK, across handoffs, across deployments, and across other frameworks entirely. An agent that learns a durable fact about a customer or a workflow records it once; the next agent, or a teammate's agent, or a Claude Code session, reads it next time. That cross-agent, cross-tool sharing is exactly what the SDK's built-in session state does not provide. Since Glen is a standard MCP server, integration is the same as adding any MCP tool, authenticated over OAuth or an API key, and from there your fleet of agents accumulates a common memory instead of each one starting cold.

FAQ

Doesn't the Agents SDK already have sessions?: Sessions persist conversation state for a given run. Glen is durable, org-shared long-term memory, read and written by every agent and MCP client in your organization, that outlives any single run or session.
How do I connect Glen to an Agents SDK agent?: Glen is a standard MCP server, so you attach it as an MCP tool the same way you would any other, authenticating over OAuth or an API key.