Agent 完成任务后提交经验 Event，生成 Capture 草稿供人工确认发布。payload 各字段对应 Fact Block，必须分别填写、内容不得重复：observation=观察到的现象/问题；decision=技术选择与理由；action=具体改动与操作；outcome=结果与验证；constraint=约束（可选）。task 仅作标题，不要复制到 observation。artifacts 须含 artifactType/artifactRole/contentRef，禁止传空对象；也可传 modifiedFiles 由服务端自动推断关联文件。
AI agents use memory_remember to create or update resources in Zhiyi — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Zhiyi environment.
| Parameter | Type | Required | Description |
|---|---|---|---|
type | string | Yes | 事件类型，如 agent_finished |
actor | string | — | Agent 标识 |
module | string | — | 模块名 |
payload | object | Yes | 经验内容，推荐分字段填写；也可传 facts 数组直接指定 Fact Blocks |
artifacts | array | — | 关联产物（修改的文件、PR 等）。每项必须填写 artifactType、artifactRole、contentRef，禁止传 {} 空对象；若不传，服务端会尝试从 action 与 modifiedFiles 推断 |
workspace | string | — | 工作空间编码（可选，默认使用 API Key 所属工作空间；勿传本地目录路径） |
repository | string | — | 代码仓库名 |
modifiedFiles | array | — | 本次任务修改过的文件路径；会自动转为 Artifact，并写入 metadata 供后端推断 |
Parameters from the server's own tool schema.
The tool writes/creates a new memory capture draft based on agent experience. It does not execute code or delete data, and requires human confirmation before publishing, making it a reversible write operation. Misuse could pollute a shared memory/knowledge base but blast radius is moderate.
From the tool's definition 生成 Capture 草稿供人工确认发布 — creates a draft capture record for human review and publication; fields include observation, decision, action, outcome, constraint, artifacts
Risk signalsAccepts raw HTML/template content (payload) · High parameter count (22 properties)
Attacks that exploit this kind of access
Agent 完成任务后提交经验 Event，生成 Capture 草稿供人工确认发布。payload 各字段对应 Fact Block，必须分别填写、内容不得重复：observation=观察到的现象/问题；decision=技术选择与理由；action=具体改动与操作；outcome=结果与验证；constraint=约束（可选）。task 仅作标题，不要复制到 observation。artifacts 须含 artifactType/artifactRole/contentRef，禁止传空对象；也可传 modifiedFiles 由服务端自动推断关联文件。. It is categorised as a Write tool in the Zhiyi MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
memory_remember accepts 8 parameters: type, actor, module, payload, artifacts, workspace, repository, modifiedFiles. Required: type, payload. The full parameter table on this page comes from the server's own tool schema.
Register the Zhiyi MCP server in PolicyLayer and add a rule for memory_remember: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Zhiyi. Nothing to install.
memory_remember is a Write tool with medium risk. Write tools should be rate-limited to prevent accidental bulk modifications.
Yes. Add a rate_limit block to the memory_remember rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for memory_remember. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
memory_remember is provided by the Zhiyi MCP server (zhiyi-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
memory_remember is one line of Zhiyi's registry record.
The record carries the whole server: verified identity, auth posture, risk grade, every tool classified, recommended policy — re-checked continuously.
Teams ship this data inside their own products. See what a licence covers →