Qoder long-term memory integration via MCP

Prerequisites

• Qoder IDE or Qoder CLI is installed and running correctly.

• Python 3.10 or later.

• The uv package manager is installed (recommended for on-demand startup with uvx).

• You have the API endpoint and authentication token for the AnalyticDB for PostgreSQL Context service.

• You have the adbpgmem_mcp_selfhosted project source code: adbpgmem-mcp-server-0.2.0.tar.gz.

  Extract the archive to a directory named adbpgmem_mcp_selfhosted. The paths in the following steps refer to this directory.

Installation

You do not need to install the MCP Server globally. uvx creates a temporary virtual environment on demand from the local project path and starts the service. You only need to ensure the source code path is accessible.

# Verify that the source code path is accessible
ls /path/to/adbpgmem_mcp_selfhosted/pyproject.toml
# This command should output the file information, confirming the path is correct

Parameters

Configuration source priority

Method 1: Inject via env (Recommended)

When registering the MCP Server, inject the environment variables directly into the env field of your settings.json file. This avoids the need to create a separate configuration file. For details, see "Register the MCP Server".

Method 2: Create a configuration file

Create a global configuration file at ~/.qoder/adbpgmem.conf or a project-level one at .qoder/adbpgmem.conf:

# Global configuration (applies to all projects)
mkdir -p ~/.qoder
cat > ~/.qoder/adbpgmem.conf << 'EOF'
ADBPGMEM_API_URL="https://api-longmemory-cn-chengdu.opentrust.net"
ADBPGMEM_API_TOKEN="sk-your-token-here"
ADBPGMEM_USER_ID="your.username"
EOF

# Project-level configuration (applies only to the current project)
mkdir -p .qoder
cat > .qoder/adbpgmem.conf << 'EOF'
ADBPGMEM_API_URL="https://api-longmemory-cn-chengdu.opentrust.net"
ADBPGMEM_API_TOKEN="sk-your-token-here"
ADBPGMEM_USER_ID="your.username"
EOF

Configuration file

The configuration file uses a KEY="VALUE" format, with one entry per line. If a configuration is missing, the MCP Server logs a warning and still starts, but subsequent API calls will fail.

Global registration (Recommended)

Edit the ~/.qoder/settings.json file. This registration applies to all your projects:

{
  "mcpServers": {
    "adbpgmem": {
      "command": "uvx",
      "args": ["--from", "/path/to/adbpgmem_mcp_selfhosted", "adbpgmem-mcp-server"],
      "env": {
        "ADBPGMEM_API_URL": "https://api-longmemory-cn-chengdu.opentrust.net",
        "ADBPGMEM_API_TOKEN": "sk-your-token-here",
        "ADBPGMEM_USER_ID": "your.username"
      }
    }
  }
}

Field description

Project-level registration

Create a .qoder/settings.json file in your project's root directory. This registration applies only to the current project and can be committed to version control for team sharing:

{
  "mcpServers": {
    "adbpgmem": {
      "command": "uvx",
      "args": ["--from", "/path/to/adbpgmem_mcp_selfhosted", "adbpgmem-mcp-server"],
      "env": {
        "ADBPGMEM_API_URL": "https://api-longmemory-cn-chengdu.opentrust.net",
        "ADBPGMEM_API_TOKEN": "sk-your-token-here",
        "ADBPGMEM_USER_ID": "your.username"
      }
    }
  }
}

Configure AGENTS.md instructions (Optional)

You can add instructions to the AGENTS.md file in your project's root directory. This file defines project-level behavioral guidelines for the Qoder AI and is loaded automatically at the start of each session. These instructions guide the AI to use memory tools proactively. The MCP tools will work without this configuration, but the AI will not use them proactively.

# MCP Servers

- **adbpgmem**: A persistent memory service. At the start of each session, use `search_memories` to look for relevant context. 
  When you find architectural decisions, debugging experiences, or coding patterns, save them using `add_memory`. 
  If memory content changes, update it with `update_memory`. It is better to record too much than to miss something.

Use the MCP tools

Tool reference

You: Remember: this project uses Go 1.22 + Gin framework, the ORM is GORM v2, and the database is PostgreSQL 15.

Qoder AI calls add_memory → The server stores the memory

You: Remember the following project decisions: We chose Kafka instead of RabbitMQ because we need the ability to replay messages.
    We selected PostgreSQL for its JSONB support. The front end uses React and TypeScript.

Qoder AI calls add_memory(infer=true) → The server extracts and stores three separate atomic facts.

You: Search for memories about the database selection.

Qoder AI calls search_memories → The server returns semantically matched historical memories.

Example use cases

You: Remember: this project uses Python 3.11 + FastAPI, the database is PostgreSQL 15, 
    the cache is Redis 7, and the message queue is Kafka 3.5.

Qoder AI calls add_memory → The tech stack information is saved to the memory store.

Session 1:
You: Remember: our APIs use the RESTful style. For pagination, use page and page_size.
    For sorting, use sort=field:asc|desc. Error responses must include the type, title, status, and detail fields.

Session 2 (a few days later):
You: Help me design a new user management API endpoint.

Qoder AI: Based on the API design standards I have in my memory, here is a design for the user management endpoint... 
         [The AI automatically follows the team's standards without you having to repeat them.]