subcog

Overview

Subcog is a sophisticated memory system designed for AI coding assistants that captures decisions, learnings, and context from coding sessions. Built in Rust, it offers a three-layer storage architecture with SQLite persistence, FTS5 indexing, and usearch HNSW vectors for semantic search. The system provides MCP server integration for seamless AI agent interoperability, supporting about 22 consolidated MCP tools for memory CRUD, search, knowledge graph operations, and prompt management. It features hybrid search combining BM25 text search with vector similarity scoring, normalized for intuitive relevance understanding.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• subcog_capture

  Capture a new memory with optional namespace and context

  Try:Capture decision to use PostgreSQL for storageTry:Log a learning about database optimization
• subcog_recall

  Search memories using hybrid semantic + BM25 search

  Try:Find memories about database decisionsTry:Search for previous solutions to authentication problems
• subcog_status

  Check the status of the memory system

  Try:Show memory system statusTry:Check memory statistics
• subcog_consolidate

  Consolidate memories to reduce redundancy

  Try:Consolidate similar memoriesTry:Optimize memory storage
• subcog_get

  Retrieve a specific memory by ID

  Try:Get memory with ID 123Try:Retrieve specific memory entry
• subcog_update

  Update an existing memory

  Try:Update memory about API design decisionsTry:Modify a previous learning
• subcog_delete

  Delete a specific memory

  Try:Delete outdated memory about deprecated methodTry:Remove irrelevant memory entry
• subcog_list

  List memories with optional filtering

  Try:List all memories in decisions namespaceTry:Show memories for current project
• subcog_graph

  Query the knowledge graph for relationships

  Try:Show connections between authentication and authorizationTry:Find related concepts
• subcog_entities

  Extract entities from memories

  Try:Extract entities from database-related memoriesTry:Identify key concepts in codebase
• subcog_prompts

  Manage prompt templates

  Try:Show available prompt templatesTry:List stored prompts
• subcog_init

  Initialize a new session with Subcog

  Try:Start new coding session with SubcogTry:Initialize Subcog for this project

Comparable tools

Installation

# Cargo (recommended for Rust developers)
cargo install subcog

# Homebrew (macOS/Linux)
brew install zircote/tap/subcog

# Docker
docker run --rm ghcr.io/zircote/subcog --help

# Binary download
curl -LO https://github.com/zircote/subcog/releases/latest/download/subcog-VERSION-TARGET.tar.gz

# npm/npx (fallback)
npx @zircote/subcog --help

{
  "mcpServers": {
    "subcog": {
      "command": "subcog",
      "args": ["serve"]
    }
  }
}

FAQ

What storage backends does Subcog support?

Subcog supports SQLite as the default backend with optional PostgreSQL support. For vector storage, it uses usearch HNSW with pgvector and Redis Vector as alternatives.

How does Subcog protect sensitive information?

Subcog includes built-in secrets detection for API keys and passwords, optional PII filtering, encryption at rest (enabled by default), and comprehensive audit logging for compliance requirements.

Compare subcog with