A sophisticated structured knowledge graph providing 18 cognitive tools for long-term agent memory, reasoning, and semantic search.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mindgraph-rs
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install mindgraph-rs using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is MindGraph?
MindGraph is a high-performance knowledge graph index designed to function as the central nervous system for AI sub-agents. It provides a structured relationship layer on top of raw files like MEMORY.md, enabling complex cross-file constraint lookups, entity resolution, and hybrid semantic search. By organizing information into specific cognitive layers—ranging from raw reality capture to intent and governance—it allows agents to maintain logical consistency across sessions.
As a core component for advanced Openclaw Skills, MindGraph transforms unstructured chat transcripts into a durable, evolving database of claims, decisions, and patterns. It supports both cloud-based API access for immediate deployment and local self-hosted servers for developers who require full control over their data and embedding processes.
MindGraph Use Cases
- Tracking complex entity relationships across multiple chat sessions and projects.
- Enforcing long-term behavioral policies and governance rules through a persistent memory layer.
- Managing high-level goals and task structures that persist beyond a single conversation.
- Building evidence-backed arguments and capturing reasoning chains for decision-making.
- Performing hybrid search (FTS + Semantic) to retrieve hyper-relevant context for sub-agents.
How MindGraph Works
- The session is initialized using the session management tool to create a unique tracking context.
- Observations, entities, and facts are ingested into the Reality Layer, which handles deduplication and resolution.
- The Epistemic Layer processes these inputs into claims, inquiries, and structures, building a logical graph of knowledge.
- Commitments and decisions are tracked through the Intent and Action layers to manage agent workflows.
- At the end of a session, a dedicated extraction pipeline flattens transcripts and uses LLMs to identify new nodes and edges.
- Maintenance scripts periodically re-index the search engine and regenerate graph-aware embeddings to ensure retrieval accuracy.
MindGraph Setup
MindGraph can be integrated into your Openclaw Skills workflow via the Cloud API or a local binary.
Cloud API (Recommended)
export MINDGRAPH_URL=https://api.mindgraph.cloud
export MINDGRAPH_TOKEN=your-api-key
Local Server
bash install.sh
bash start.sh
export OPENAI_API_KEY=sk-your-key-for-embeddings
MindGraph Data Schema & Taxonomy
MindGraph utilizes a comprehensive taxonomy to organize agent knowledge across several functional layers:
| Layer | Key Node Types | Functional Purpose |
|---|---|---|
| Reality | Entity, Source, Observation | Captures raw data and unique identifiers for people/orgs. |
| Epistemic | Claim, Argument, Structure | Models reasoning, evidence, and conceptual patterns. |
| Intent | Goal, Decision, Commitment | Tracks what the agent intends to do and why. |
| Memory | Session, Journal, Summary | Manages the lifecycle of conversations and their distillation. |
| Agent | Task, Plan, Policy | Governs execution steps and behavioral constraints. |
MindGraph Advanced Features
- Entity Resolution Pipeline: A 5-step process using fuzzy matching and semantic similarity to prevent duplicate nodes.
- Hybrid Retrieval: Uses Reciprocal Rank Fusion (RRF) to combine the strengths of Full-Text Search and vector embeddings.
- Toulmin Argumentation: Specifically designed nodes for claims, evidence, and warrants to improve agent reasoning.
- Automated Extraction: Multi-model pipeline that selects between Gemini Flash and Pro based on transcript size for optimal data ingestion.
- Deep-Merge Props: Support for complex metadata updates that allow agents to evolve existing knowledge without overwriting history.
SKILL.md
Loading
Related Openclaw Skills
Prompt Request Pipeline
A fully automated GitHub pipeline that transforms issues tagged with [auto] into tested, reviewed, and merged pull requests.
Openclaw Recovery Codex
A comprehensive diagnostic and recovery agent for Openclaw infrastructure, ensuring Gateway, channels, and pipelines remain operational across all platforms.
Multi-User Workspace Management
A robust framework for managing multiple users, permissions, and isolated sandboxes within an AI agent environment.
Mission Control
A Next.js-powered command center for managing OpenClaw tasks, memories, and team activities through a visual web interface.
Stock Analyzer
A powerful tool for fetching Yahoo Finance data, analyzing portfolios, and summarizing earnings reports for US and Japanese markets.
English Poetry Hub
A collaborative multiplayer game skill for AI agents to write, critique, and finalize English poetry in real-time.
METADATA
Featured*
