MCP prompts

The TekMemo MCP server exposes two built-in prompts that guide the LLM toward grounded memory operations.

Prompt table

Name	Title	Purpose
tekmemo-recall-context	Recall TekMemo Context	Turn a user question into a grounded recall instruction
tekmemo-memory-review	Review Memory Candidate	Review whether text should become durable TekMemo memory

tekmemo-recall-context

Guides the model to use TekMemo recall to answer a user question, returning grounded context before the answer.

Arguments:

Name	Type	Required	Description
query	string	Yes	The user question or task
workspaceId	string	No	Optional workspace ID
includeGraph	boolean	No	Whether to use graph-aware recall (default: true)

Generated prompt:

Use TekMemo recall to answer this request.
Query: <query>
Workspace: <workspaceId or "runtime default">
Graph-aware recall: <yes/no>

Return grounded context first, then the answer.

tekmemo-memory-review

Guides the model to review and classify candidate memory before writing it to TekMemo.

Arguments:

Name	Type	Required	Description
content	string	Yes	Potential memory content (max 100,000 chars)
workspaceId	string	No	Optional workspace ID

Generated prompt:

Review this candidate memory before writing it to TekMemo.
Workspace: <workspaceId or "runtime default">

Candidate:
<content>

Classify it as durable, short-lived, sensitive, redundant, or unsafe.
Explain whether it should be saved.

Example usage

# In an MCP client, invoke a prompt:
# tekmemo-recall-context with query="What's the sync strategy?"
# tekmemo-memory-review with content="Use D1 for sync truth."

Prompts are exposed via the prompts/list and prompts/get MCP protocol methods. Any MCP host can list them and invoke them with arguments.