anchored

README

Anchored

Persistent cross-tool memory for AI coding agents. Single binary. Zero dependencies.

Anchored is an MCP memory server that gives Claude Code, Cursor, OpenCode, and any MCP-compatible tool a shared, persistent memory. Install once — all your tools read, write, and search the same knowledge base.

No API keys. No daemon. All embeddings run locally.

---

Install

curl -fsSL https://github.com/jholhewres/anchored/install.sh | bash

From source (Go 1.24+):

git clone https://github.com/jholhewres/anchored.git
cd anchored && make build && ./bin/anchored init

Setup

anchored init          # auto-detects Claude Code, Cursor, OpenCode and registers MCP
anchored import --all  # imports existing sessions and memories

That's it. Open any tool — Anchored is available.

Usage

From your AI tool (via MCP)

Anchored registers as MCP tools automatically. Your agent can:

• anchored_context — load relevant memory for the current project
• anchored_search — search across all memories (semantic + keyword)
• anchored_save — persist a fact, decision, or preference
• anchored_list — list memories by category or project
• anchored_forget — remove outdated memories
• kg_query / kg_add — query and build a knowledge graph

From the terminal

anchored search "deploy gateway"       # search memories
anchored save "API runs on :8080"      # save a fact
anchored list --category decision      # list decisions
anchored context                      # show context for current directory
anchored stats                        # memory overview
anchored import claude-code           # import from a specific source
anchored identity edit                 # edit your preferences

What it imports

Source	Extracts
Claude Code	JSONL sessions, memory/ files, CLAUDE.md
OpenCode	opencode.db sessions, messages, todos
Cursor	.cursor/rules/
DevClaw	data/memory.db full memory + knowledge graph
Any directory	Markdown, code, docs

All content is sanitized (secrets removed) before storage.

What you end up with

~/.anchored/
├── bin/anchored           # the binary
├── data/
│   ├── anchored.db        # SQLite (FTS5 + vectors + knowledge graph)
│   └── onnx/              # local embedding model (~33MB)
└── config.yaml

No daemon. No ports. The binary runs on demand via MCP STDIO — your tool manages the lifecycle.

Docs

• Design — memory stack, hybrid search, knowledge graph, quantization
• Architecture — project structure and implementation details
• Import Sources — how each tool's data is parsed

License

MIT