# TencentDB Agent Memory Supported Agent Frameworks: Complete Integration Guide

> Discover TencentDB Agent Memory's supported agent frameworks including Claude Code, DeepSeek Harness, Codex and more. Seamlessly integrate with zero-code proxy architecture.

- Repository: [Tencent Cloud/TencentDB-Agent-Memory](https://github.com/TencentCloud/TencentDB-Agent-Memory)
- Tags: getting-started
- Published: 2026-08-21

---

**TencentDB Agent Memory supports seven major agent frameworks natively—including Claude Code, DeepSeek Harness, Codex, CodeBuddy, WorkBuddy, Hermes, and OpenClaw—plus any generic OpenAI-compatible client, through a zero-code proxy architecture that automatically injects Chat Memory, Skills, Wiki, and CodeGraph assets without requiring any changes to agent source code.**

TencentDB Agent Memory is an open-source memory-server designed to augment LLM-based coding agents with persistent team knowledge. The TencentCloud/TencentDB-Agent-Memory repository provides native integration for popular frameworks via a lightweight **Memory Proxy** that intercepts standard LLM API calls and injects memory assets transparently.

## Native Agent Framework Support

According to the agents table in [`README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/README.md), TencentDB Agent Memory offers **direct support** for the following frameworks:

- **DeepSeek Harness** – Configure the proxy as documented in the "Using Proxy with DeepSeek Harness" section
- **Claude Code** – Point the client at the proxy endpoint with standard environment variables
- **Codex** – Available via the "Using Proxy with Codex" integration guide
- **CodeBuddy** – Uses dedicated "Using Proxy with CodeBuddy" instructions
- **WorkBuddy** – Enterprise agent support documented under "Using Proxy with WorkBuddy"
- **Hermes** – Proxy-compatible Hermes agent configuration
- **OpenClaw** – Direct connection support for OpenClaw agents

All supported agents consume the same **memory-server** backend, ensuring consistent access to **Chat Memory**, **Skills**, **Wiki**, and **CodeGraph** assets across every framework.

## Zero-Code Proxy Architecture

The integration relies on the **Memory Proxy** located in `MemoryProxy/`, a lightweight forwarding layer that runs on port `8096` by default. As detailed in [`MemoryProxy/README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/MemoryProxy/README.md), the proxy:

1. Intercepts normal LLM API calls from the agent
2. Injects team memory assets into the request context
3. Forwards to the upstream LLM provider
4. Handles session initialization and conversation write-back automatically

This design ensures **zero-code integration**—you only change the base URL and authentication token that the agent uses to call the LLM, with no modifications to agent source code required.

## Generic Framework Integration

For agents not listed in the native support table, TencentDB Agent Memory provides a **generic integration guide**. Any LLM agent that uses standard OpenAI or Anthropic client libraries can connect by pointing its base URL to the proxy endpoint.

According to the "Using Proxy with Other Platforms – Generic" section in [`README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/README.md), this approach works with custom OpenAI clients, proprietary agent frameworks, and experimental LLM tools.

## Configuration Examples

The [`INSTALL.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/INSTALL.md) file provides concrete configuration steps for each supported framework, including step 3 which illustrates the exact commands for Claude Code.

### Claude Code Setup

Configure Claude Code to route through the memory proxy using environment variables:

```bash

# Set the proxy endpoint (default port 8096)

export ANTHROPIC_BASE_URL=http://127.0.0.1:8096/claude-code/default

# Use the business-user key generated by the admin panel

export ANTHROPIC_AUTH_TOKEN="sk-mem-<32-char-key>"

# Run Claude Code normally; memory injection happens automatically

claude --model <your-upstream-model>

```

### Generic OpenAI Client Setup

For custom agents using OpenAI-compatible APIs:

```bash

# Redirect API calls to the memory proxy

export OPENAI_API_BASE=https://127.0.0.1:8096/v1

# Authenticate with the memory server user key

export OPENAI_API_KEY="sk-mem-<32-char-key>"

# Execute any standard OpenAI client script

python my_openai_script.py

```

## Key Implementation Files

The following source files define the supported integrations and proxy behavior:

| File | Description |
|------|-------------|
| [`README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/README.md) | Overview, supported agents table, and high-level architecture |
| [`INSTALL.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/INSTALL.md) | Step-by-step deployment and proxy configuration examples |
| [`MemoryProxy/README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/MemoryProxy/README.md) | Proxy forwarding logic and agent binding mechanisms |
| [`MemoryCore/README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/MemoryCore/README.md) | Core memory service APIs (Chat Memory, Skills, Wiki, CodeGraph) |
| [`deploy/global-images/README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/deploy/global-images/README.md) | Deployment scripts and environment-variable templates |

These files collectively define how each agent framework connects to the shared memory backend through the proxy layer.

## Summary

- TencentDB Agent Memory supports **seven native agent frameworks**: DeepSeek Harness, Claude Code, Codex, CodeBuddy, WorkBuddy, Hermes, and OpenClaw
- Integration requires **zero code changes**—only endpoint and token configuration via environment variables
- The **Memory Proxy** (`MemoryProxy/`) runs on port `8096` and automatically injects Chat Memory, Skills, Wiki, and CodeGraph assets
- **Generic integration** allows any OpenAI-compatible or Anthropic-compatible client to connect by changing base URLs
- Configuration details are documented in [`README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/README.md), [`INSTALL.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/INSTALL.md), and the [`MemoryProxy/README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/MemoryProxy/README.md) directory

## Frequently Asked Questions

### Which agent frameworks are supported natively by TencentDB Agent Memory?

TencentDB Agent Memory provides native integration for DeepSeek Harness, Claude Code, Codex, CodeBuddy, WorkBuddy, Hermes, and OpenClaw. These frameworks connect through the Memory Proxy without requiring source code modifications—only environment variable changes to point at the proxy endpoint on port `8096`.

### How do I integrate a custom agent framework not listed in the documentation?

Use the generic integration pattern documented in [`README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/README.md) under "Using Proxy with Other Platforms – Generic." Any agent that uses standard OpenAI (`OPENAI_API_BASE`) or Anthropic (`ANTHROPIC_BASE_URL`) environment variables can connect by redirecting these to the memory proxy URL `http://127.0.0.1:8096/v1` or the appropriate framework-specific path.

### What memory assets are injected into agent requests?

Every connected agent gains access to four core asset types managed by `MemoryCore/`: **Chat Memory** (conversation history), **Skills** (reusable code snippets), **Wiki** (team documentation), and **CodeGraph** (repository structure and relationships). The proxy injects these automatically without code changes.

### Where are the specific configuration commands documented?

Concrete proxy configuration commands for each framework are located in [`INSTALL.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/INSTALL.md), specifically step 3 for Claude Code and corresponding sections for other agents. The [`MemoryProxy/README.md`](https://github.com/TencentCloud/TencentDB-Agent-Memory/blob/main/MemoryProxy/README.md) file details the proxy's role in forwarding LLM requests and binding agents to memory assets.