# Installation & Setup Guide Reviewed for v12.6.0 on 2026-11-05. Python 3.11 or newer is required. Python 3.10 support ended in v12.5.0. ## Local installation Create and activate a Python 3.11+ virtual environment, then install: ```bash python -m pip install mcp-memory-service memory server --help ``` The default backend is `sqlite_vec`; local ONNX embeddings do not require a cloud API key. PyTorch and transformers are optional extras rather than prerequisites for the default installation. ## Connect an MCP client For Claude Code: ```bash claude mcp add memory -- memory server ``` For Claude Desktop, use the configuration paths and JSON example in the [Integration Guide](03-Integration-Guide). Restart the client and check for `memory_store` and `memory_search`. ## HTTP API and dashboard For a local development instance: ```bash MCP_ALLOW_ANONYMOUS_ACCESS=true memory server --http ``` In another terminal: ```bash curl --fail http://127.0.0.1:8000/api/health ``` Open the dashboard at `http://127.0.0.1:8000` and the REST reference at `http://127.0.0.1:8000/api/docs`. HTTP is the local default; HTTPS requires certificates or a configured reverse proxy. For shared or remote deployments, configure API-key or OAuth authentication and HTTPS instead of anonymous access. ## Choose a deployment - [Setup Guide](https://github.com/doobidoo/mcp-memory-service/blob/main/docs/setup-guide.md): MCP only, OpenCode, Codex hooks, dashboard, cloud sync and team setups. - [Storage Backends](https://github.com/doobidoo/mcp-memory-service/blob/main/docs/guides/STORAGE_BACKENDS.md): SQLite-vec, Hybrid, Cloudflare and Milvus. Hybrid requires Cloudflare configuration; selecting it alone is not a complete setup. - [Docker configurations](https://github.com/doobidoo/mcp-memory-service/tree/main/tools/docker): Compose files and Dockerfiles live in `tools/docker/`. - [Production Guide](https://github.com/doobidoo/mcp-memory-service/blob/main/docs/deployment/production-guide.md): deployment and authentication. - [systemd Service](https://github.com/doobidoo/mcp-memory-service/blob/main/docs/deployment/systemd-service.md): Linux service setup. - [External Embeddings](https://github.com/doobidoo/mcp-memory-service/blob/main/docs/deployment/external-embeddings.md): Ollama, vLLM, TEI and OpenAI-compatible endpoints. ## Upgrading Use Python 3.11+ and back up the database before upgrading: ```bash python -m pip install --upgrade mcp-memory-service ``` Restart the service using your deployment's service manager. Review the [release notes](https://github.com/doobidoo/mcp-memory-service/releases) and [migration guide](https://github.com/doobidoo/mcp-memory-service/blob/main/docs/MIGRATION.md), especially when upgrading from v10 or earlier: deprecated MCP tool aliases were removed in v11.