sediment-mcp 0.2.2

[![Crates.io](https://img.shields.io/crates/v/sediment-mcp.svg)](https://crates.io/crates/sediment-mcp)
[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
[![CI](https://github.com/rendro/sediment/actions/workflows/ci.yml/badge.svg)](https://github.com/rendro/sediment/actions/workflows/ci.yml)

# Sediment

Semantic memory for AI agents. Local-first, MCP-native.

Combines vector search, a relationship graph, and access tracking into a unified memory intelligence layer — all running locally as a single binary.

## Why Sediment?

- **Single binary, zero config** — no Docker, no Postgres, no Qdrant. Just `sediment`.
- **Sub-25ms recall** — local embeddings and vector search, no network round-trips.
- **5-tool focused API** — `store`, `recall`, `list`, `forget`, `connections`. That's it.
- **Works everywhere** — macOS (Intel + ARM), Linux x86_64. All data stays on your machine.

### Comparison

| | Sediment | OpenMemory MCP | mcp-memory-service |
|---|---|---|---|
| Install | Single binary | Docker + Postgres + Qdrant | Python + pip |
| Dependencies | None | 3 services | Python runtime + deps |
| Tools | 5 | 10+ | 24 |
| Embeddings | Local (all-MiniLM-L6-v2) | API-dependent | API-dependent |
| Graph features | Built-in | No | No |
| Memory decay | Built-in | No | No |

## Install

```bash
# Via crates.io
cargo install sediment-mcp

# Via Homebrew
brew tap rendro/tap
brew install sediment

# Via shell installer
curl -fsSL https://raw.githubusercontent.com/rendro/sediment/main/install.sh | sh

# From source
cargo install --path .
```

## Setup

Add Sediment to your MCP client configuration:

### Claude Desktop

Add to `~/Library/Application Support/Claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "sediment": {
      "command": "sediment"
    }
  }
}
```

### Claude Code

Run `sediment init` in your project, or add manually to `~/.claude/settings.json`:

```json
{
  "mcpServers": {
    "sediment": {
      "command": "sediment"
    }
  }
}
```

### Cursor

Add to `.cursor/mcp.json` in your project:

```json
{
  "mcpServers": {
    "sediment": {
      "command": "sediment"
    }
  }
}
```

### VS Code (Copilot)

Add to `.vscode/mcp.json` in your project:

```json
{
  "servers": {
    "sediment": {
      "command": "sediment"
    }
  }
}
```

### Windsurf

Add to `~/.codeium/windsurf/mcp_config.json`:

```json
{
  "mcpServers": {
    "sediment": {
      "command": "sediment"
    }
  }
}
```

### JetBrains IDEs

Go to **Settings > Tools > AI Assistant > MCP Servers**, click **+**, and add:

```json
{
  "sediment": {
    "command": "sediment"
  }
}
```

## Tools

| Tool | Description |
|------|-------------|
| `store` | Save content with optional title, tags, metadata, expiration, scope, replace, and related item links |
| `recall` | Search memories by semantic similarity with decay scoring, trust weighting, graph expansion, and co-access suggestions |
| `list` | List stored items by scope (project/global/all) with tag filtering |
| `forget` | Delete an item by ID (removes from vector store and graph) |
| `connections` | Show relationship graph for an item (related, supersedes, co-accessed edges) |

## CLI

```bash
sediment           # Start MCP server
sediment init      # Set up Claude Code integration
sediment stats     # Show database statistics
sediment list      # List stored items
```

## How It Works

### Three-Database Hybrid

All local, embedded, zero config:

- **LanceDB** — Vector embeddings and semantic similarity search
- **SQLite** (graph) — Relationship tracking: RELATED, SUPERSEDES, CO_ACCESSED, CLUSTER_SIBLING edges
- **SQLite** (access) — Mutable counters: access tracking, decay scoring, consolidation queue

### Key Features

- **Memory decay**: Results re-ranked by freshness (30-day half-life) and access frequency. Old memories rank lower but are never auto-deleted.
- **Trust-weighted scoring**: Validated and well-connected memories score higher.
- **Project scoping**: Automatic context isolation between projects. Same-project items get a similarity boost.
- **Relationship graph**: Items linked via RELATED, SUPERSEDES, and CO_ACCESSED edges. Recall expands results with 1-hop graph neighbors and co-access suggestions.
- **Background consolidation**: Near-duplicates (≥0.95 similarity) auto-merged; similar items (0.85–0.95) linked.
- **Auto-tagging**: Items without tags inherit tags from similar existing items.
- **Type-aware chunking**: Intelligent splitting for markdown, code, JSON, YAML, and plain text.
- **Conflict detection**: Items with ≥0.85 similarity flagged on store.
- **Cross-project recall**: Results from other projects flagged with provenance metadata.
- **Local embeddings**: all-MiniLM-L6-v2 via Candle (384-dim vectors, no API keys).

## Performance

Sub-25ms recall latency at 10K items with full graph features enabled. See [BENCHMARKS.md](BENCHMARKS.md) for detailed numbers.

| DB Size | Graph Off | Graph On |
|---------|-----------|----------|
| 100     | ~8ms      | ~10ms    |
| 1,000   | ~12ms     | ~15ms   |
| 10,000  | ~18ms     | ~23ms   |

### Data Location

- Vector store: `~/.sediment/data/`
- Graph + access tracking: `~/.sediment/access.db`

Everything runs locally. Your data never leaves your machine.

## Contributing

See [CONTRIBUTING.md](CONTRIBUTING.md) for build instructions and PR guidelines.

## License

[MIT](LICENSE)