TencentDB Agent Memory Service Ports: Complete Network Configuration Guide

TencentDB Agent Memory exposes five distinct TCP ports across its microservices architecture: 8420 for the MemoryCore gateway, 8096 for MemoryProxy, 8123 for the MemoryPanel UI, 8421 for MemoryKnowledge, and 8125/8424 for the combined deployment image.

The TencentDB Agent Memory platform operates as a distributed system of containerized microservices, with each component binding to specific TCP ports for HTTP communication. According to the TencentCloud/TencentDB-Agent-Memory repository, these port assignments are hardcoded in the respective Dockerfiles and environment variables, ensuring consistent networking behavior across Docker and Kubernetes deployments.

Default Port Assignments by Service

MemoryCore Gateway (Port 8420)

The core gateway API serves as the primary entry point for the platform. In MemoryCore/Dockerfile, the service exposes port 8420 as the default TDAI_GATEWAY_PORT. This port handles all incoming API requests and routes them to appropriate backend services.

MemoryProxy (Port 8096)

Acting as an intermediary layer, MemoryProxy forwards client requests to the core gateway. The MemoryProxy/Dockerfile defines port 8096 for this service, enabling load balancing and request routing before traffic reaches the gateway.

MemoryPanel UI (Port 8123)

The front-end web interface provides visual interaction with the memory service. Built from MemoryPanel/docker/local/Dockerfile.local, this React-based UI exposes port 8123 for browser access.

MemoryKnowledge Service (Port 8421)

Handling embeddings and vector search operations, the knowledge-graph service runs on port 8421 as specified in MemoryKnowledge/Dockerfile. This port serves RAG (Retrieval-Augmented Generation) queries and semantic search requests.

Combined Deployment (Ports 8125 and 8424)

For simplified deployments, the repository provides a unified image combining both UI and Knowledge services. The deploy/panel-knowledge-combined/Dockerfile exposes two ports: 8125 (PANEL_PORT) for the web interface and 8424 (KNOWLEDGE_PORT) for the knowledge-graph API.

Port Configuration in Source Files

Each service binds to 0.0.0.0 inside its container, making the ports accessible from the host network. Health checks utilize these same endpoints—for example, the UI container validates readiness via curl http://127.0.0.1:${PANEL_PORT}/health.

The port definitions are centralized in the following build artifacts:

• MemoryCore/Dockerfile — Exposes 8420
• MemoryProxy/Dockerfile — Exposes 8096
• MemoryPanel/docker/local/Dockerfile.local — Exposes 8123
• MemoryKnowledge/Dockerfile — Exposes 8421
• deploy/panel-knowledge-combined/Dockerfile — Exposes 8125 and 8424

Docker Deployment Examples

To run individual services with their default ports exposed:

# MemoryCore gateway on port 8420

docker run -d -p 8420:8420 tencentdb-agent-memory:core

# MemoryProxy on port 8096

docker run -d -p 8096:8096 tencentdb-agent-memory:proxy

# MemoryPanel UI on port 8123

docker run -d -p 8123:8123 tencentdb-agent-memory:panel

# MemoryKnowledge on port 8421

docker run -d -p 8421:8421 tencentdb-agent-memory:knowledge

For the combined deployment exposing both UI and Knowledge ports:

docker run -d \
  -p 8125:8125 \
  -p 8424:8424 \
  tencentdb-agent-memory:combined

Summary

• Port 8420: MemoryCore gateway API (TDAI_GATEWAY_PORT)
• Port 8096: MemoryProxy request forwarding layer
• Port 8123: MemoryPanel web UI
• Port 8421: MemoryKnowledge embeddings service
• Ports 8125/8424: Combined image for UI (PANEL_PORT) and Knowledge (KNOWLEDGE_PORT)
• All services listen on 0.0.0.0 and are defined in their respective Dockerfiles

Frequently Asked Questions

What is the default port for the TencentDB Agent Memory gateway?

The MemoryCore gateway service defaults to port 8420, defined by the TDAI_GATEWAY_PORT environment variable in MemoryCore/Dockerfile. This port serves as the primary API entry point for the entire platform.

Can I change the default ports when running the containers?

Yes, you can map the container ports to different host ports using Docker's -p flag (e.g., -p 8080:8420), though the internal container ports remain fixed by the EXPOSE directives in the Dockerfiles. For permanent changes, you would need to modify the environment variables like PANEL_PORT or KNOWLEDGE_PORT in the combined Dockerfile and rebuild the image.

Which ports are required for the combined Panel-Knowledge deployment?

The combined image requires two ports: 8125 for the web UI (MemoryPanel) and 8424 for the knowledge-graph API (MemoryKnowledge). These are defined in deploy/panel-knowledge-combined/Dockerfile through the PANEL_PORT and KNOWLEDGE_PORT environment variables.

How do I verify that the services are running on the correct ports?

Each service exposes a health check endpoint on its designated port. For example, the UI responds to http://127.0.0.1:8123/health, while other services expose similar endpoints on their respective ports (8420, 8096, 8421). You can verify connectivity using curl or by checking the container logs for successful binding messages.

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