Prerequisites for TencentDB Agent Memory: Complete Setup Guide

To deploy TencentDB Agent Memory, you need Docker and Docker Compose installed, a Git client to clone the repository, a reachable LLM endpoint with valid API credentials, and four open network ports (8420, 8125, 8424, 8096) on your host machine.

TencentDB Agent Memory is an open-source agent memory platform maintained by Tencent Cloud that enables persistent context for AI coding agents. Before you can run the start-all.sh script to launch the Memory Core, Panel UI, Knowledge Service, and Proxy containers, you must satisfy several infrastructure and configuration requirements documented throughout the repository.

Core Infrastructure Requirements

The stack is delivered as a containerized architecture defined in the deploy/global-images/ directory. You cannot run the services natively without Docker.

Docker and Docker Compose

You must have the docker command available on your system. The installation guide in INSTALL.md (lines 20-27) documents the "Full three-in-one stack" deployment that pulls and starts all containers simultaneously. This includes the Memory Core, Memory Hub Panel, Knowledge Service, and Proxy services.

Git Client

A Git installation is required to clone the repository before executing the startup scripts. The same section in INSTALL.md (lines 21-24) shows the git clone command sequence necessary to obtain the source files and deployment scripts.

LLM Configuration and API Credentials

The system requires a valid Large Language Model endpoint to function. The Memory Core and Proxy services forward requests to upstream providers like Anthropic, OpenAI, or compatible services.

Environment Variables

You must provide the base URL, API key, and model name using specific environment variable prefixes. As shown in MemoryProxy/.env.example, the system expects:

• MEMORY_* variables for the internal memory hub configuration
• PROXY_* variables for the external proxy configuration

The interactive deploy/global-images/start-all.sh script (lines 31-35) walks you through configuring these variables during the first run. If an .env file does not exist, the script automatically copies .env.example to .env and prompts for your credentials.

Runtime Configuration

The MemoryCore/src/utils/env.ts file contains the runtime code that reads these environment variables and configures the services accordingly. Ensure your LLM endpoint is reachable from the host machine before starting the stack.

Network and Port Configuration

The services bind to fixed ports that must be available and reachable on your host machine. According to INSTALL.md (lines 57-62), the default port allocation is:

• 8420: Memory Core service
• 8125: Memory Hub Panel UI (configured in MemoryPanel/web/vite.config.ts)
• 8424: Knowledge Service (defined in MemoryKnowledge/Dockerfile)
• 8096: Proxy service

Verify these ports are not occupied by other applications before running ./start-all.sh.

Authentication Setup

On first boot, the deployment scripts generate administrative credentials required to access the system.

Admin Key Generation

The start-all.sh script creates an admin user_key automatically during initialization. This key is saved to deploy/global-images/.admin-key as documented in the installation guide (lines 43-46). You need this key to log into the Panel UI at http://localhost:8125 and to create regular business users for your agents.

User Management

After accessing the Panel UI with the admin key, you can generate additional user_key values for specific agent clients like Claude Code, Codex, or DeepSeek Harness. The list of supported agents and their configuration locations appears in INSTALL.md (lines 78-89).

Optional Components

You can deploy components individually if you do not require the full stack.

Standalone Knowledge Service

If you only need the Memory Hub UI without the Knowledge Service, you can run containers separately. The "Memory Hub only" section in INSTALL.md (lines 58-70) provides the specific Docker run command. This configuration still requires the LLM environment variables and a REMOTE_INSTANCE_URL pointing to the running Memory Core instance on port 8420.

Quick Start Commands

Once prerequisites are satisfied, use these commands to launch the system:

# Clone the repository

git clone https://github.com/TencentCloud/TencentDB-Agent-Memory.git
cd TencentDB-Agent-Memory/deploy/global-images

# Run the interactive setup script

./start-all.sh

The script validates connectivity, generates the admin key, and launches all containers. After completion, access the Panel UI:

open http://localhost:8125

# Use the admin key printed in the terminal output

To configure Claude Code or another agent to use your local proxy:

export ANTHROPIC_BASE_URL=http://127.0.0.1:8096/claude-code/default
export ANTHROPIC_AUTH_TOKEN="<business_user_key>"

Summary

• Docker and Docker Compose are mandatory for running the containerized stack defined in the deploy/global-images/ directory.
• Git is required to clone the TencentCloud/TencentDB-Agent-Memory repository and access the start-all.sh script.
• LLM credentials must be configured via MEMORY_* and PROXY_* environment variables in the .env file, auto-generated from .env.example.
• Network ports 8420, 8125, 8424, and 8096 must be open on the host machine for the Core, Panel, Knowledge, and Proxy services respectively.
• Admin authentication relies on an auto-generated key stored in deploy/global-images/.admin-key, created on first run as documented in INSTALL.md (lines 43-46).

Frequently Asked Questions

Can I run TencentDB Agent Memory without Docker?

No. The architecture relies on containerization for the Memory Core, Panel UI, Knowledge Service, and Proxy components. The start-all.sh script and all installation paths in INSTALL.md assume Docker and Docker Compose are installed and available on the host system.

Which LLM providers are compatible with the system?

The platform supports any Anthropic, OpenAI, or OpenAI-compatible API endpoint. You configure the base URL, API key, and model name using the MEMORY_* and PROXY_* variable prefixes in the .env file. The start-all.sh script validates connectivity to your specified endpoint before starting the containers.

What happens if port 8096 is already in use?

The services use fixed ports defined in the source configuration. If port 8096 (Proxy), 8420 (Memory Core), 8125 (Panel), or 8424 (Knowledge) are occupied, the containers will fail to bind. You must free these specific ports or modify the port mappings in the Docker configuration files before running ./start-all.sh.

How do I recover the admin key if I lose it?

The admin user_key is generated once during the first execution of start-all.sh and stored in deploy/global-images/.admin-key. If you lose this key, you must delete the .admin-key file and restart the stack to generate a new administrative credential. This process is described in INSTALL.md (lines 43-46).

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}