Is TencentDB Agent Memory Open Source? MIT License and Architecture Breakdown

Yes, TencentDB Agent Memory is fully open-source software released under the permissive MIT license, granting unrestricted rights to use, modify, and distribute the code.

TencentDB Agent Memory is an open-source project hosted at TencentCloud/TencentDB-Agent-Memory that provides a shared memory layer for LLM-driven agents. The repository includes complete source code, documentation, and deployment scripts, all freely available under the MIT license. This shared memory system enables agents to persist conversation history, skills, and knowledge across sessions using a layered L0-L3 architecture.

MIT License Confirmation

The project includes an explicit LICENSE file containing the full MIT license text. According to the source code, this license grants anyone the right to use, copy, modify, merge, publish, distribute, sublicense, and sell the software. The README.md displays a license badge that links directly to this file, confirming the open-source status.

Architecture Overview

TencentDB Agent Memory implements a distributed memory system through three core microservices.

MemoryCore

MemoryCore serves as the central hub storing and serving memory assets including Chat Memory, Skills, Wiki, and CodeGraph. It handles asset versioning, access control lists (ACLs), and team management. Key documentation resides in MemoryCore/v3-api-memorycore-doc.md.

MemoryKnowledge

MemoryKnowledge provides HTTP APIs for querying Wiki and CodeGraph assets with strict ID validation. The service implements standardized response envelopes in MemoryKnowledge/src/api-helpers.ts.

MemoryProxy

MemoryProxy acts as a language-model-agnostic proxy that injects memory assets into LLM requests. It supports WorkBuddy, Claude Code, and other frameworks without code changes. The proxy implements request classification, session handling, asset injection, and Langfuse observability in MemoryProxy/src/workbuddyHandler.ts.

Deploying the Open Source Stack

The repository includes one-click deployment scripts in deploy/global-images/start-all.sh that launch memory-core, memory-hub, and proxy containers.

Clone and start the services:

# Clone the repo

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

# Copy the example env file and edit the LLM parameters

cp .env.example .env
$EDITOR .env   # set the two sets of LLM credentials

# Launch all services (MemoryCore, MemoryHub, Proxy) in one command

./start-all.sh

After execution, the management UI is available at http://localhost:8125.

Integration Examples

Node.js Client Calling MemoryProxy

const fetch = require('node-fetch');

async function callWorkBuddy(prompt) {
  const response = await fetch('http://localhost:8125/workbuddy/v1/chat/completions', {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      // Replace with your API key if required
      'Authorization': 'Bearer <YOUR_API_KEY>',
    },
    body: JSON.stringify({
      model: 'gpt-4o-mini',
      stream: false,
      input: [
        {
          type: 'message',
          role: 'user',
          content: [{ type: 'input_text', text: prompt }],
        },
      ],
    }),
  });

  const data = await response.json();
  console.log('Agent reply:', data);
}

callWorkBuddy('Explain the purpose of the Wiki asset.');

The proxy automatically injects relevant memory assets—Chat Memory, Skills, and Wiki pages—before forwarding requests upstream.

Querying Wiki via REST API

curl -X POST https://your-host/v3/wiki/query \
     -H "Content-Type: application/json" \
     -H "x-tdai-service-id: my_service" \
     -d '{
           "team_id": "team123",
           "query": "What is the onboarding process?",
           "top_k": 5
         }'

The response follows the envelope defined in api-helpers.ts, guaranteeing a code, message, and data payload.

Data Model and Security

The system processes memory through layered stages: L0 (raw conversation), L1 (facts), L2 (scenario), and L3 (persona). Assets are bound to agents via ACLs supporting private, team, and restricted access levels, enabling secure sharing across different frameworks without code modifications.

Summary

• TencentDB Agent Memory is confirmed open-source under the MIT license via the LICENSE file and README badge.
• The architecture consists of three microservices: MemoryCore, MemoryKnowledge, and MemoryProxy.
• Complete source code is available in the TencentCloud/TencentDB-Agent-Memory repository with no proprietary dependencies.
• Deployment requires only Docker and a single script: deploy/global-images/start-all.sh.
• The system supports multiple LLM frameworks through a language-agnostic proxy with automatic asset injection.

Frequently Asked Questions

What license is TencentDB Agent Memory released under?

TencentDB Agent Memory is released under the MIT license. The full license text is available in the LICENSE file at the repository root, granting permissions to use, copy, modify, merge, publish, distribute, sublicense, and sell copies of the software.

Can I modify and redistribute TencentDB Agent Memory commercially?

Yes. The MIT license explicitly permits commercial use, modification, and distribution. Organizations can fork the repository, customize the memory services, and deploy them in production environments without royalty fees or source code disclosure requirements.

Where can I find the source code for the memory services?

The source code is publicly available in the TencentCloud/TencentDB-Agent-Memory GitHub repository. Key components include MemoryCore/ for the central hub, MemoryKnowledge/src/api-helpers.ts for API utilities, and MemoryProxy/src/workbuddyHandler.ts for proxy logic.

How do I deploy the open source TencentDB Agent Memory stack?

Clone the repository and execute deploy/global-images/start-all.sh after configuring your environment variables. This script launches all required services including MemoryCore, MemoryHub, and the Proxy, making the system accessible at localhost:8125.

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