Description

Save and search personal notes, decisions, and insights with semantic search via MCP. Use this skill when users want to remember things across conversations,...

README (SKILL.md)

KnowMine — Personal Knowledge Base for AI Agents

Name: KnowMine
Author: yiing99

Save notes, decisions, and insights. Search them later by meaning, not keywords.

KnowMine is a remote MCP server that gives your AI agent a personal knowledge base with semantic search. Your agent can save knowledge entries during conversations and retrieve them later using natural language queries. All data is stored per-user with full isolation.

Setup

1. Get your API key

Sign up at knowmine.ai, go to Settings → MCP, and copy your key.

2. Connect

Remote MCP server — no Docker or local install needed.

Server URL:  https://knowmine.ai/api/mcp
Transport:   Streamable HTTP
Auth header: Authorization: Bearer \x3CYOUR_KNOWMINE_API_KEY>

For OpenClaw:

npx clawhub@latest install knowmine

Tools (11)

Knowledge CRUD

Tool	What it does
`add_knowledge`	Save a note, insight, dev log, article, or reflection. Auto-generates title and tags.
`search_my_knowledge`	Search your knowledge base by meaning. Supports type and tag filters.
`get_knowledge`	Get full content of one entry by ID.
`update_knowledge`	Edit title, content, type, or tags of an entry.
`delete_knowledge`	Remove an entry.
`get_related_knowledge`	Find entries related to a given entry by semantic similarity.
`list_folders`	List your folders and recent entries in each.

Memory

Tool	What it does
`save_memory`	Save a decision, lesson, insight, or preference from a conversation.
`recall_memory`	Search past memories by natural language. Filter by type.

Analysis

Tool	What it does
`get_soul`	Generate a user profile summary based on your knowledge. Exportable as a system prompt.
`get_insight`	Analyze your knowledge for patterns — frequent topics, recurring themes.

Example Usage

Save a decision during work:

save_memory: "Chose PostgreSQL over MongoDB for the new project — need ACID transactions and our team knows SQL well"

Find it later in a new session:

recall_memory: "database decision for new project"
→ Returns the PostgreSQL decision with full context

Search across all knowledge:

search_my_knowledge: "deployment strategies"
→ Returns matching notes ranked by relevance, even if you never used the word "deployment" in the original entry

How It Works

Search: pgvector semantic similarity on PostgreSQL (not keyword matching)
Embeddings: OpenAI text-embedding-3-small, auto-generated on save
Protocol: MCP over Streamable HTTP
Auth: API key via Bearer token
Data isolation: Each API key maps to one user account
Network: All requests go to knowmine.ai only (no third-party data sharing)
Rate limit: 60 requests/minute

Source Code

GitHub: github.com/YIING99/knowmine
Website: knowmine.ai
Health check: knowmine.ai/api/mcp/health

Built by YIING99

Usage Guidance

Before installing: 1) Confirm the metadata mismatch — ask the publisher why the registry listed no env vars but SKILL.md requires KNOWMINE_API_KEY. 2) Clarify the data flow for embeddings: does KnowMine send your content to OpenAI (or other third parties) to compute embeddings? If so, what is their privacy policy and can you opt out or use a self‑hosted embedding provider? 3) Verify the website and GitHub repo links, check the /api/mcp/health endpoint over HTTPS, and prefer testing with non‑sensitive dummy notes first. 4) Ensure the API key can be rotated/revoked and has limited scope; do not store secrets you wouldn’t want sent to a third party. 5) If you need a stricter privacy model, ask whether self‑hosting or an alternative that keeps embeddings local is available. If the publisher cannot clearly answer these points, treat the skill as higher risk.

Capability Analysis

Type: OpenClaw Skill Name: knowmine Version: 1.2.0 The 'knowmine' skill is a remote Model Context Protocol (MCP) implementation for a personal knowledge base hosted at knowmine.ai. Its functionality is limited to standard CRUD operations for notes and semantic search (e.g., add_knowledge, search_my_knowledge, save_memory). While it requires network access and transmits user-provided data to a third-party service, this behavior is explicitly documented and aligned with its stated purpose as a cloud-based memory tool. No evidence of malicious intent, prompt injection, or unauthorized data exfiltration was found in SKILL.md or _meta.json.

Capability Assessment

ℹ Purpose & Capability

The declared purpose (personal semantic notes via MCP) matches the runtime instructions and the single credential (KNOWMINE_API_KEY) the SKILL.md says is required. However the registry metadata earlier listed no required env vars while the SKILL.md demands KNOWMINE_API_KEY — this metadata mismatch is unexpected and should be clarified.

⚠ Instruction Scope

Instructions restrict network endpoints to knowmine.ai and describe the MCP API and tools, which is appropriate. But the SKILL.md also states 'Embeddings: OpenAI text-embedding-3-small' and elsewhere claims 'no third‑party data sharing' — these statements conflict: if embeddings are generated via OpenAI, user content may be forwarded to OpenAI, which is a third party. The skill gives no explicit guidance about what user data is sent to embedding providers or whether embeddings are computed entirely server-side under KnowMine's control.

✓ Install Mechanism

This is an instruction-only skill with no install spec or shipped code, so there is nothing written to disk by the skill itself. SKILL.md suggests using 'npx clawhub@latest install knowmine' (an external install command) but no downloads or extracts are embedded in the skill — install risk is low from the skill bundle itself.

⚠ Credentials

SKILL.md requires a single service credential (KNOWMINE_API_KEY), which is proportional for a hosted personal-knowledge service. But the registry metadata omitted this requirement (incoherent). Also the mention of OpenAI embeddings raises the question whether an OpenAI key or other credentials are needed or whether KnowMine will forward user content to OpenAI on the backend — that data flow is not documented and could increase credential/privilege needs or privacy exposure.

✓ Persistence & Privilege

The skill does not request always-on presence and uses default autonomous invocation settings. It does not declare modifications to other skills or system-wide settings. No elevated persistence privileges are requested in the provided metadata or instructions.

Version History

v1.2.0

v1.2.0: Add permissions declaration, rewrite description for clarity, remove sensitive wording that triggered false positive moderation flag, add source code link

v1.1.0

v1.1.0: Added folder management, knowledge association discovery, AI-powered insights, improved semantic search with pgvector

v1.0.0

Initial release - AI Agent Personal Knowledge Base MCP Server

Metadata

Slug knowmine

Version 1.2.0

License MIT-0

All-time Installs 0

Active Installs 0

Total Versions 3

Frequently Asked Questions

What is KnowMine?

Save and search personal notes, decisions, and insights with semantic search via MCP. Use this skill when users want to remember things across conversations,... It is an AI Agent Skill for Claude Code / OpenClaw, with 347 downloads so far.

How do I install KnowMine?

Run "/install knowmine" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is KnowMine free?

Yes, KnowMine is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does KnowMine support?

KnowMine is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created KnowMine?

It is built and maintained by YING99 (@yiing99); the current version is v1.2.0.

More Skills

KnowMine