phloem

Phloem is an MCP-compatible server that provides persistent memory for your AI across sessions. It stores memories locally in a SQLite database and exposes them through the MCP protocol over standard input/output, so any MCP-capable client can connect without needing a network port. Once running, tools that support MCP—such as Claude Code, VS Code integrations, Cursor, Windsurf, Zed, Neovim, Cline, Warp, Continue, JetBrains, and others—can read and write memories, citations, and causal graph data to maintain context as you work. Start the binary and connect an MCP client; the server will handle semantic search, memory linking to specific code locations, and automatic decay of memories as code evolves.

To use Phloem with tools, run the phloem executable on your machine (for example after installing via brew or downloading the binary) and then configure your MCP clients to connect via the MCP interface. For Claude Code or other supported tools, there are setup commands or guides that auto-detect environments and bind to the local Phloem instance. The memory data is stored locally (in ~/.phloem) and can be audited with phloem audit, ensuring offline operation with no external network dependencies.

Prerequisites:

• A supported environment (macOS, Windows, or Linux)
• Go 1.24+ if building from source
• A C compiler if building sqlite-vec CGO components (often provided by your system)

Installation options:

1. Install via Homebrew (macOS):

brew install canopyhq/tap/phloem
phloem setup

2. Download binary (Windows / Linux / macOS):

• Go to GitHub Releases for CanopyHQ/phloem and download the appropriate binary for your platform, extract it, add to PATH, then run:

phloem setup

3) Build from source:
```bash
git clone https://github.com/CanopyHQ/phloem.git && cd phloem
CGO_ENABLED=1 go build -o phloem .
./phloem setup

Notes:

• Phloem requires Go 1.24+ and a C compiler for sqlite-vec when building from source.
• The setup command auto-detects tools on your machine and configures memory capture for MCP-compatible clients.

Tips and common considerations:

• Phloem operates offline with no network surface; all memories are stored locally in ~/.phloem/memories.db and accessed via JSON-RPC over stdio as defined by MCP.
• You can audit memory usage and verify there is no network activity using phloem audit and system checks.
• If you upgrade Phloem, memories persist across versions, but always verify compatibility with your MCP clients.
• For privacy, there is no telemetry by default; you can delete memories with rm -rf ~/.phloem at any time.
• If you encounter build issues related to sqlite-vec, ensure you have a C compiler and proper CGO tooling installed on your platform.