Skip to content
Docs

Optional Features

CodeForge includes several features that are available but not enabled by default. This page covers how to enable and configure them.

Adds persistent vector memory to Claude Code via a Qdrant MCP server. Claude can store and retrieve information across sessions using qdrant-store and qdrant-find tools.

Add to devcontainer.json under "features":

"./features/mcp-qdrant": {
"collectionName": "my-project-memory",
"embeddingModel": "all-MiniLM-L6-v2"
}
OptionDefaultDescription
collectionNameagent-memoryQdrant collection name
embeddingModelall-MiniLM-L6-v2Embedding model for vector search
qdrantUrl(empty)Remote Qdrant server URL. If empty, uses local storage.
qdrantApiKey(empty)API key for remote Qdrant server
qdrantLocalPath/workspaces/.qdrant/storageLocal storage path (when no URL set)
  • all-MiniLM-L6-v2 — default, smallest, fastest
  • BAAI/bge-small-en-v1.5
  • BAAI/bge-base-en-v1.5
  • sentence-transformers/all-mpnet-base-v2
  1. During container build, the embedding model is pre-downloaded from GCS (not HuggingFace, to avoid network issues in containers).
  2. On container start, a post-start hook registers the Qdrant MCP server in Claude Code’s settings.json.
  3. Claude Code can then use qdrant-store and qdrant-find tools to persist and search memories.
Terminal window
uvx mcp-server-qdrant --help

Any feature can be disabled without removing it from devcontainer.json by setting "version": "none":

"./features/hadolint": { "version": "none" },
"./features/shellcheck": { "version": "none" }

The feature entry stays in the config for easy re-enabling — just remove "version": "none" or set it to "latest".

  • Configuration — settings.json and devcontainer.json reference
  • Plugins — plugin system overview and per-plugin configuration
  • Troubleshooting — solutions for build and feature issues