Frequently Asked Questions

TencentDB Agent Memory is a team-level memory hub for AI Agents that transforms conversations, documents, and code into reusable memory assets. It creates four key memory types: Chat Memory, Skills, LLM-Wiki, and CodeGraph. These assets are governed, shared, and equipped across different agents and frameworks, enabling teams to accumulate and reuse knowledge effectively.

By extracting reusable assets from previous work, Agent Memory eliminates the need to repeatedly explain project context, reread documents, or rediscover workflows. Chat Memory retains user preferences and interaction history, Skills store reusable workflows, Wiki structures knowledge, and CodeGraph indexes code relationships. This creates a persistent knowledge base that new agents can immediately utilize.

• Chat Memory: Retains user preferences, facts, decisions, and interaction history across sessions
• Skills: Reusable workflows extracted from conversations with versions, boundaries, and validation rules
• LLM-Wiki: Structured documentation pages with link graphs for product docs and specs
• CodeGraph: Indexed code symbols, files, call relationships, and impact paths

Installation is simple with our one-command setup:

git clone https://github.com/Tencent/TencentDB-Agent-Memory.git
cd TencentDB-Agent-Memory/deploy/global-images
cp .env.example .env
# Edit .env with your LLM parameters
./start-all.sh

This launches all services and provides a URL for Claude integration. Access the dashboard at http://localhost:8125.

TencentDB Agent Memory currently integrates with:

• OpenClaw (v2026.3.13+)
• Hermes Agent Gateway
• Claude Code
• CodeBuddy
• SDK for custom integrations
• Framework adapters are in active development

Memory assets have four visibility levels:

• Private: Only visible to the asset owner
• Team: Visible to all team members
• Restricted: Precise access via User/Role/Agent ACLs
• Agent: For targeted equipping of specific agents

Two role layers exist: System Admins manage users and teams, while Team Admins manage assets within their team.

The system has three core components:

• Memory Core: Processes and stores memory assets in layered structure (L0-L3)
• Memory Hub: Web interface for managing teams, assets, and agent loadouts
• Proxy: Integrates with external agents and frameworks

Memory grows through conversation distillation and knowledge indexing, enabling efficient retrieval based on context and permissions.

We welcome contributions in several areas:

• Report bugs and issues on GitHub
• Suggest features in GitHub Discussions
• Submit pull requests with code improvements
• Help improve documentation
• Share benchmark results
• Develop framework integrations

Join our Discord community to connect with core developers.

TencentDB Agent Memory is licensed under the MIT License. This permits free use, modification, and distribution of the software, provided all copyright notices and disclaimer are included. See the full license here.