# Agent memory

Source: https://www.gryz.ai/topics/agent-memory/
Updated: 2026-11-03

Agent memory is the part of shared context that holds the durable facts, preferences and decisions your AI agents save, so they stop re-learning the same things every chat.

**Short answer:** Agent memory keeps what your agents learn after the chat ends. In Gryz it sits next to your documents, tasks, skills and search, so every agent shares one store across Claude, Cursor, ChatGPT and any MCP app. You can see who saved each fact, then edit, export or forget it.

## Start here

- [Agent Memory overview](https://www.gryz.ai/docs/agent-memory/overview/): Give AI assistants durable, cross-session facts about you and your projects, scoped and quota-bound per plan.
- [Agent Memory terms](https://www.gryz.ai/docs/agent-memory/terms/): Plain-English meanings for Agent Memory labels you see in Gryz: kinds, importance, working set, JSONL, and MCP.
- [Get your agent ready for Gryz](https://www.gryz.ai/docs/getting-started/agents/): Run the Gryz agent onboarding router after connecting an MCP client, with safe local-write consent and a portable manual fallback.
- [MCP tools reference](https://www.gryz.ai/docs/mcp/tools/): Inputs, behavior, safety annotations and limits for every Gryz MCP tool: memory, Notes, documents, Projects, Tasks, Skills, Search and company connections.
- [Memory API reference](https://www.gryz.ai/docs/api/memory/): REST API for creating and listing Agent Memory records, mirroring the remember and list_memories MCP tools.

## Read more

- [How do I share context between Claude Code and Cursor?](https://www.gryz.ai/gryzle/share-context-between-claude-code-and-cursor/): Connect Claude Code and Cursor to Gryz and they share one context for your repo: decisions, docs, tasks and skills. Here's the prompt to paste.
- [Gryz vs ChatGPT memory: what's the difference?](https://www.gryz.ai/gryzle/gryz-vs-chatgpt-memory/): ChatGPT memory is great inside ChatGPT. Gryz is shared context for every agent you use: memory, docs with comments and versions, tasks, skills and search.
- [Gryz vs Claude Projects and memory: when do you need both?](https://www.gryz.ai/gryzle/gryz-vs-claude-projects/): A Claude Projects alternative, or a partner? Gryz is shared context that travels from Claude to Claude Code, Cursor and ChatGPT: memory, docs and search.
- [Gryz vs Mem0: which one should you use?](https://www.gryz.ai/gryzle/gryz-vs-mem0/): Looking for a Mem0 alternative? Mem0 is a memory API for apps you build. Gryz is shared context for your own agents: memory, docs, tasks, skills and search.
- [The Four Kinds of Memory Your Agent Actually Needs](https://www.gryz.ai/gryzle/types-of-agent-memory/): 'Memory' sounds like one thing. It isn't. Here's how a fact, a preference, an episode and a procedure differ, and why your agent needs to know which is which.
- [Agent memory that works across every model, and that you own](https://www.gryz.ai/gryzle/portable-agent-memory/): Use the best model for each job and switch as the leaderboard moves. That only works when your memory travels with you instead of sitting in one assistant.

## FAQ [#faq]

**Is agent memory the same as shared context?**
No. Memory is one part of it. [Shared context](/topics/shared-context/) also holds your documents, tasks, skills and notes, with one search across all of them.

**What kinds of memory does an agent need?**
Four: facts, preferences, episodes and procedures. Each one ages differently, so treating them all the same is how memory breaks. [This post](/gryzle/types-of-agent-memory/) walks through each kind.

**How do agents save and find memories?**
After you [connect Gryz](/docs/mcp/setup/), an agent saves a fact with `remember` and looks facts up with `recall`. `list_memories` browses what is saved, and `forget` takes a fact back.

**Can Claude and Cursor use the same memory?**
Yes. Any MCP client reads and writes the same store. Personal memories follow you everywhere, and project memories are keyed to a repo or workspace.

**Who can see my memories?**
You can. Every memory shows who saved it, and you can edit, export or forget it. Your memories stay private to your account. Gryz never uses them to train models and never shows them to anyone else. In a company workspace, shared memory follows the company's permissions.

**Is agent memory free?**
You can start with a free Gryz account. Each plan sets how many memories and projects you can keep. See the [plan limits](/docs/api/memory/#plan-limits).