mnemo provides cloud-persistent, multi-agent memory for AI agents using TiDB Serverless to enable cross-session recall and hybrid vector search.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mnemo-memory
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 mnemo-memory 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 mnemo-memory?
mnemo is a high-performance memory layer designed to solve the problem of session amnesia in AI agents. While most agents are stateless, mnemo externalizes their memory into TiDB Cloud Serverless, ensuring that context, decisions, and learned data persist indefinitely. This allows developers to maintain continuity across sessions and even share intelligence between different agent platforms.
By integrating mnemo into your workflow, you enhance your Openclaw Skills with the ability to perform semantic and keyword searches across past interactions. It treats memory as a shared resource, breaking down silos between tools like Claude Code and OpenCode, and uses advanced conflict resolution to handle simultaneous updates from multiple agents.
mnemo-memory Use Cases
- Persisting agent context across multiple restarts to eliminate session amnesia.
- Sharing a unified memory pool between different agents like Claude Code and OpenCode.
- Synchronizing agent knowledge across different physical machines and devices.
- Implementing team-wide shared intelligence with space isolation and per-agent tokens.
- Managing concurrent memory updates using CRDT-based conflict resolution.
How mnemo-memory Works
- The AI agent generates a memory or stores a specific context using the
memory_storetool. - The mnemo plugin assigns a CRDT vector clock to the data to manage versioning and concurrency.
- Data is transmitted via HTTP API to a TiDB Cloud Serverless instance.
- If configured, TiDB generates embeddings server-side using the
EMBED_TEXT()function, removing the need for external API keys. - When recall is needed, the agent uses
memory_searchto perform a hybrid vector and keyword query. - The system merges and ranks results, providing the agent with the most relevant historical context.
mnemo-memory Setup
To install the plugin for use with your Openclaw Skills, run the following command:
npm install mnemo-openclaw
Add the plugin configuration to your openclaw.json file. You can obtain your TiDB credentials by creating a free cluster at tidbcloud.com:
{
"plugins": {
"slots": { "memory": "mnemo" },
"entries": {
"mnemo": {
"enabled": true,
"config": {
"host": "<your-tidb-host>",
"username": "<your-tidb-user>",
"password": "<your-tidb-pass>"
}
}
}
}
}
mnemo-memory Data Schema & Taxonomy
mnemo organizes data within TiDB using a schema optimized for both relational metadata and vector-based semantic search:
| Component | Type | Description |
|---|---|---|
| Memory ID | String | Unique identifier for the memory record. |
| Content | Text | The primary text data or context stored by the agent. |
| Vector | VECTOR | Native TiDB vector type for ANN cosine similarity search. |
| Vector Clock | JSON | CRDT-based clock used to resolve multi-agent write conflicts. |
| Metadata | JSON | Additional attributes such as agent ID, session tags, or timestamps. |
| Tombstone | Boolean | Used for soft-deletes to maintain CRDT integrity. |
mnemo-memory Advanced Features
- Hybrid Search Engine: Automatically combines vector-based semantic search with traditional SQL keyword matching for maximum precision.
- Zero-Ops Auto-Embeddings: Leverages TiDB's server-side
EMBED_TEXT()function to generate vectors without requiring an OpenAI API key. - CRDT Conflict Resolution: Uses deterministic tie-breaking and vector clocks to allow multiple agents to write to the same memory pool without data loss.
- Team Mode Support: Supports a dedicated server mode that provides space isolation and per-agent token authentication for collaborative environments.
- Soft Deletion: Ensures that deleted memories do not "resurrect" in multi-agent environments by using tombstone records.
SKILL.md
Loading
Related Openclaw Skills
IP Lookup
A network utility skill for retrieving public IP addresses and detailed geolocation information via CLI commands.
Dream of Clawra
An automated media-sharing skill that selects and sends contextually relevant images and dance videos to messaging channels via OpenClaw.
Whisper Speech-to-Text (STT)
An AI-powered speech-to-text skill that utilizes OpenAI Whisper to transcribe audio files into accurate text automatically.
Session Memory Skill
A skill that automatically archives AI conversations into dated Markdown files while masking sensitive data for secure retrieval.
Brave Search & Proxy Configuration
A comprehensive guide to setting up Brave Search API and resolving network connectivity or proxy issues for Openclaw web tools.
YouTube Long Video Transcript & Translation
A specialized workflow for extracting, translating, and summarizing verbatim transcripts from long YouTube videos using automated agent orchestration.
Featured*
