DecisionNode

Overview

DecisionNode provides a structured memory store that allows developers to document technical decisions with proper context and rationale. These decisions are stored as JSON with embeddings, making them semantically searchable. The system offers both CLI commands and MCP tools for interacting with the decision store. It features conflict detection, history tracking, deprecation/activation of decisions, and includes a local web UI for visualization. The server can be integrated with Claude Code, Cursor, Windsurf, and other MCP clients.

Try asking AI

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

• add_decision

  Add a new decision to the structured memory store

  Try:Add a decision: 'Use PostgreSQL for production database'Try:Record decision: Implement CI/CD pipeline
• search_decisions

  Semantically search for decisions in the memory store

  Try:Find decisions related to database scalingTry:Search for decisions about authentication implementation
• update_decision

  Update an existing decision in the memory store

  Try:Update decision backend-007 with new constraintsTry:Modify the rationale for decision ui-003
• delete_decision

  Delete a decision from the memory store

  Try:Remove decision frontend-012Try:Delete outdated decision about monorepo structure
• list_decisions

  List all decisions in the memory store

  Try:Show all decisions in Backend scopeTry:List all active decisions
• decision_history

  Get the history of changes for a decision

  Try:Show history for decision backend-007Try:Who modified decision auth-001?
• deprecate_decision

  Deprecate a decision (soft delete)

  Try:Deprecate decision old-auth-systemTry:Mark decision monolith-001 as deprecated
• activate_decision

  Activate a deprecated decision

  Try:Activate decision legacy-featureTry:Bring back decision api-v1
• check_conflicts

  Check for conflicts before adding a new decision

  Try:Check conflicts for new decision about cachingTry:Verify if similar decision already exists

Comparable tools

Installation

npm install -g decisionnode
cd your-project
decide init      # creates project store
decide setup     # configure Gemini API key (free tier)

# Connect to Claude Code (run once)
claude mcp add decisionnode -s user decide-mcp

{
  "mcpServers": {
    "decisionnode": {
      "command": "npx",
      "args": ["decisionnode", "mcp"]
    }
  }
}

FAQ

How are decisions stored?

Decisions are stored as JSON files with metadata including id, scope, decision text, status, rationale, constraints, and timestamps. Each decision is also embedded as a vector using Gemini's embedding model for semantic search.

Can I use DecisionNode without an API key?

No, you need to set up a Gemini API key during initialization. The free tier is sufficient for most use cases, as only embedding functionality requires the API.

On Hacker News

Recent discussion from the developer community.

• Show HN: DecisionNode – shared structured memory for all AI coding tools via MCP▲ 23💬 5
  Story by AmmarSaleh50 · 2026-04-10

Compare DecisionNode with