---
title: "Copilot Studio agent memory: how it works"
description: "Copilot Studio agents can remember things per user: free-form memory on the Microsoft Copilot harness, a fixed schema for customer service. What and how long."
image: "https://bombos.ai/assets/og-en-v2.png"
---

# Do Copilot Studio agents have memory?

Short answer

Yes, in preview. On the Microsoft Copilot harness a maker can switch on memory, and the agent then keeps a private store per user that is deleted after 28 days without interaction. On the Customer Conversation harness memory follows a fixed schema with name, open intents and a summary, kept for 30 days. Neither shares anything between users.

*Piet Baudoin · published November 17, 2026 · November 17, 2026*

## Two kinds of memory, depending on the harness

In Copilot Studio the “harness” is the runtime under your agent. Memory works differently on the two harnesses that support it (documentation as of November 2026, both in preview):

|  | Microsoft Copilot harness | Customer Conversation harness |
| --- | --- | --- |
| Use case | Employee agents that reason and work in steps | Customer service over chat and voice |
| What is stored | Free-form facts the agent captures from the conversation | A fixed schema: customer name, list of past open intents, summary of the last conversation, plus up to 5 custom parameters |
| Scope | Per user, per agent | Per customer, per agent; across channels through a shared contact identifier |
| Retention | Deleted after 28 days without interaction | 30 days |
| Who sees it | Only the user | Never shared with other customers |
| Switched on by | The maker, with the **Memory** toggle on the **Build** tab | The maker, per agent |

Microsoft documents memory only for these two harnesses, not for the standard harness with rule-based topics.

## How memory works on the Microsoft Copilot harness

Microsoft describes three steps:

1.  **Capture**: during a conversation the agent picks out facts worth keeping, such as a preference or a role.
2.  **Store**: these go to a separate store for that user and that agent, in Microsoft-managed storage.
3.  **Apply**: in later conversations the agent uses them to tailor its answers.

The user stays in control. They can ask the agent in chat what it remembers or to forget something, and the memory portal has **View memories** and **Delete all memories**. Memories expire after 28 days without interaction. Memory is disabled in group chats and Teams channels.

## How memory works for customer service

The Customer Conversation harness uses structured memory because a support conversation needs predictable fields, not loose notes. Besides the three fixed fields, a maker can add up to five custom parameters, each with a description of 30 to 300 characters and up to 10 allowed values. Each field can be set to **Replace** (keep only the latest value) or **Append** (add to a list). When the same contact identifier is used, memory carries across channels, so a customer who chatted on the website and later calls is recognized.

## The common mistake: thinking memory makes the agent learn

Memory makes an agent remember a person, not learn a job. What user A tells the agent is invisible to user B, and to the maker. If a team lead corrects the agent (“we never quote prices by email”), that correction lives only in the team lead’s private memory store, and is deleted after 28 days without interaction. Rules that should apply to everyone still belong in the instructions or the knowledge sources, edited by a maker. The same split applies to Microsoft 365 Copilot itself, see [Does Microsoft 365 Copilot have memory?](https://bombos.ai/en/questions/copilot/copilot-memory/).

## What this means for a team

1.  Turn memory on for agents where personal preferences save time: a writing helper, an onboarding agent.
2.  Keep team rules out of memory. Write them into the instructions, and decide who maintains them.
3.  Tell users memories expire after 28 days of not using the agent, so occasional users will start over.
4.  For customer service, define the custom parameters before go-live: they are the only structured fields you get.

For more on putting company context into agents in general, see [How to give AI agents your company context](https://bombos.ai/en/knowledge-base/how-to-give-ai-agents-your-company-context/).

## Where Bombos fits

Copilot Studio memory is per user, per agent and short-lived by design; what the team learns does not travel. Bombos is a plugin that gives Copilot, your Studio agents and the other AI tools your team uses (Claude, ChatGPT, Codex, Cursor, your own agents) one shared memory: team knowledge, learned rules, mail and documents, traceable, in the cloud or on your own server. One person can try it first with their own agents.

## Related questions

Can the maker see what an agent remembers about a user?

No. Microsoft states memories on the Microsoft Copilot harness are private to the user. The user can view and delete them through the agent's memory settings, the maker can't.

Does turning off memory delete what was stored?

No. Switching the Memory toggle off stops the agent from using memory, but existing memories are not deleted by it. They expire after 28 days without interaction or when the user deletes them.

Does memory work in Teams channels?

No. Memory is disabled in group chats and Teams channels, so an agent there treats every conversation without stored memories.

Can the agent learn something from one user and use it for another?

No. Both memory types are per user per agent and never shared. Knowledge for everyone still goes in the agent's instructions and knowledge sources.

## Sources

1.  [Microsoft Learn: Memory in agents (Microsoft Copilot harness)](https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/memory-overview)
2.  [Microsoft Learn: Memory for the Customer Conversation harness](https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/memory-overview-customer-conversation)
3.  [Microsoft Learn: Harnesses in Copilot Studio](https://learn.microsoft.com/en-us/microsoft-copilot-studio/harnesses-overview)

## What does your AI forget?

1.  01You tell us which AI you use and where it keeps starting over.
2.  02We show what a shared memory changes there, on your own examples.
3.  03You keep the plan, with or without Bombos.

![Piet Baudoin](https://bombos.ai/assets/team/piet-klein.jpg)

**Piet Baudoin**You talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe

## More questions about Microsoft Copilot

-   [What is the Copilot Studio Kit, now the Copilot Agent Kit?](https://bombos.ai/en/questions/copilot/copilot-studio-kit/)
-   [How do you govern Copilot Studio agents as an admin?](https://bombos.ai/en/questions/copilot/copilot-studio-admin-governance/)
-   [What are workflows and agent flows in Copilot Studio?](https://bombos.ai/en/questions/copilot/copilot-studio-agent-flows/)
-   [How do you publish a Copilot Studio agent to Microsoft Teams?](https://bombos.ai/en/questions/copilot/publish-copilot-studio-agent-to-teams/)
-   [What is Copilot Studio lite, now Agent Builder?](https://bombos.ai/en/questions/copilot/what-is-copilot-studio-lite/)
-   [How does Copilot Studio pricing work: credits, capacity packs and pay-as-you-go?](https://bombos.ai/en/questions/copilot/copilot-studio-agent-pricing/)
-   [What license do you need for Copilot Studio?](https://bombos.ai/en/questions/copilot/copilot-studio-license/)
-   [How do you connect an MCP server to Copilot Studio?](https://bombos.ai/en/questions/copilot/copilot-studio-mcp/)

[All questions about Microsoft Copilot →](https://bombos.ai/en/questions/copilot/)


---

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://bombos.ai/#organization","name":"Bombos","url":"https://bombos.ai/","logo":"https://bombos.ai/assets/bombos-bee.png","description":"Bombos is the memory layer under every AI your team uses. One shared memory for Claude, ChatGPT, Copilot, Cursor and your own agents, that learns how you work.","sameAs":["https://www.linkedin.com/company/bombos-ai","https://www.youtube.com/@truebombos","https://www.instagram.com/truebombos/","https://bsky.app/profile/bombostrue.bsky.social","https://www.threads.com/@truebombos","https://mastodon.social/@Bombos","https://www.tiktok.com/@truebombos","https://t.me/truebombos","https://www.pinterest.com/truebombos/","https://github.com/bombos-ai","https://linktr.ee/bombos.ai","https://bombosai.substack.com","https://hashnode.com/@bombosai","https://hackernoon.com/u/bombos","https://dev.to/bombos"],"email":"info@bombos.ai","areaServed":["NL","Europe"],"knowsLanguage":["nl","en"],"contactPoint":{"@type":"ContactPoint","email":"info@bombos.ai","contactType":"sales","url":"https://bombos.ai/en/contact/","availableLanguage":["nl","en"]}},{"@type":"WebSite","@id":"https://bombos.ai/#website","name":"Bombos","url":"https://bombos.ai/","inLanguage":["nl-NL","en"],"publisher":{"@id":"https://bombos.ai/#organization"}}]}

{"@context":"https://schema.org","@type":"Article","headline":"Do Copilot Studio agents have memory?","description":"Copilot Studio agents can remember things per user: free-form memory on the Microsoft Copilot harness, a fixed schema for customer service. What and how long.","abstract":"Yes, in preview. On the Microsoft Copilot harness a maker can switch on memory, and the agent then keeps a private store per user that is deleted after 28 days without interaction. On the Customer Conversation harness memory follows a fixed schema with name, open intents and a summary, kept for 30 days. Neither shares anything between users.","datePublished":"2026-11-17T00:00:00.000Z","dateModified":"2026-11-17T00:00:00.000Z","author":{"@type":"Person","name":"Piet Baudoin"},"publisher":{"@type":"Organization","@id":"https://bombos.ai/#organization","name":"Bombos","logo":{"@type":"ImageObject","url":"https://bombos.ai/assets/bombos-bee.png"}},"image":"https://bombos.ai/assets/og-en-v2.png","mainEntityOfPage":"https://bombos.ai/en/questions/copilot/copilot-studio-agent-memory/","inLanguage":"en-US","isPartOf":{"@type":"CollectionPage","name":"Microsoft Copilot","url":"https://bombos.ai/en/questions/copilot/"},"citation":[{"@type":"CreativeWork","name":"Microsoft Learn: Memory in agents (Microsoft Copilot harness)","url":"https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/memory-overview"},{"@type":"CreativeWork","name":"Microsoft Learn: Memory for the Customer Conversation harness","url":"https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/memory-overview-customer-conversation"},{"@type":"CreativeWork","name":"Microsoft Learn: Harnesses in Copilot Studio","url":"https://learn.microsoft.com/en-us/microsoft-copilot-studio/harnesses-overview"}]}

{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Bombos","item":"https://bombos.ai/en/"},{"@type":"ListItem","position":2,"name":"Questions","item":"https://bombos.ai/en/questions/"},{"@type":"ListItem","position":3,"name":"Microsoft Copilot","item":"https://bombos.ai/en/questions/copilot/"},{"@type":"ListItem","position":4,"name":"Do Copilot Studio agents have memory?","item":"https://bombos.ai/en/questions/copilot/copilot-studio-agent-memory/"}]}

{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"Can the maker see what an agent remembers about a user?","acceptedAnswer":{"@type":"Answer","text":"No. Microsoft states memories on the Microsoft Copilot harness are private to the user. The user can view and delete them through the agent's memory settings, the maker can't."}},{"@type":"Question","name":"Does turning off memory delete what was stored?","acceptedAnswer":{"@type":"Answer","text":"No. Switching the Memory toggle off stops the agent from using memory, but existing memories are not deleted by it. They expire after 28 days without interaction or when the user deletes them."}},{"@type":"Question","name":"Does memory work in Teams channels?","acceptedAnswer":{"@type":"Answer","text":"No. Memory is disabled in group chats and Teams channels, so an agent there treats every conversation without stored memories."}},{"@type":"Question","name":"Can the agent learn something from one user and use it for another?","acceptedAnswer":{"@type":"Answer","text":"No. Both memory types are per user per agent and never shared. Knowledge for everyone still goes in the agent's instructions and knowledge sources."}}]}
```
