MCP server that records development decisions as structured JSON, embeds them as vectors, and enables semantic search over past decisions.
MCP server that records development decisions as structured JSON, embeds them as vectors, and enables semantic search over past decisions.
Decision Node is a Developer Tools MCP server that records development decisions as structured JSON, embeds them as vectors, and lets AI coding tools semantically search prior decisions, with the DecisionNode software offered free while Gemini API usage may incur separate usage-based costs. It is built for developers and teams using Claude Code, Cursor, Windsurf, Antigravity, or any MCP client who want persistent project memory without relying on chat history.
DecisionNode focuses on a narrow but important workflow: preserving architecture, product, UI, and implementation decisions so future AI coding sessions can retrieve them before generating code. The website describes a decision flow where a choice such as “use our custom dropdown, not the native one” is stored as a typed JSON object with fields like id, scope, decision, rationale, and constraints. That record is then embedded with Gemini’s gemini-embedding-001 model and stored locally in vectors.json. When a later AI session needs context, it can call the search_decisions MCP tool, embed the new query, compare it with stored vectors using cosine similarity, and return the closest matches with similarity scores. The example on the site shows a 94% match for a prior UI decision.
The tool is not just a simple note-taking layer. It includes a CLI, an MCP server, local storage, global decisions that apply across projects, history tracking, conflict detection, and a local web UI launched with decide ui. The UI includes three specific views: a force-directed graph of similar decisions, a 2D vector space using UMAP projection of 3072-dimensional embeddings, and a searchable list. The website also states that matched nodes pulse live in the color of the MCP client that searched, which makes retrieval activity visible while Claude Code, Cursor, Windsurf, Antigravity, or another MCP client is working.
Compared to the 870+ AI tools in our directory, DecisionNode is unusually specific: it is not a broad AI coding assistant, documentation platform, or generic vector database. It is best understood as a shared structured memory layer for AI-assisted development. Teams should consider it when repeated decisions are being lost between AI sessions, when ADR-style notes are too disconnected from the coding workflow, or when multiple AI coding clients need to search the same local memory. It is less suitable for teams that need hosted collaboration, provider choice for embeddings, or deep integrations with existing documentation tools such as ADR repositories, Notion, or internal knowledge bases, because the website emphasizes local vectors.json storage, Gemini embeddings, and MCP-based retrieval rather than cloud synchronization.
Was this helpful?
DecisionNode stores decisions as typed JSON objects rather than unstructured notes. The website example includes id, scope, decision, rationale, and constraints, which gives future AI sessions enough context to understand both the choice and why it was made.
AI coding tools can call the search_decisions MCP tool to retrieve relevant prior decisions. Search is based on vector similarity, so a future query about a settings dropdown can surface an earlier decision about using a custom selector even if the wording differs.
DecisionNode uses Gemini’s gemini-embedding-001 model to convert decision text into vectors. The website states that these embeddings are 3072-dimensional and stored locally in vectors.json.
Running decide ui launches a local interface with three views: a force-directed graph, a 2D vector space using UMAP projection, and a searchable list. The site also says matched nodes pulse live in the searching client’s color, making AI retrieval activity visible.
DecisionNode includes configurable similarity thresholds, global decisions, history tracking, conflict detection, and deprecate/reactivate workflows. Conflict detection checks existing decisions at 75% similarity before adding a new one, which helps catch duplicates and contradictions.
$0/month
Ready to get started with Decision Node?
View Pricing Options →We believe in transparent reviews. Here's what Decision Node doesn't handle well:
Weekly insights on the latest AI tools, features, and trends delivered to your inbox.
No-Code & Automation
All-in-one workspace that combines notes, databases, wikis, project management, and AI-powered writing into a flexible block-based platform for individuals and teams
Knowledge & Documents
AI workspace for knowledge management and team collaboration from Atlassian.
AI agent memory
Memory infrastructure for AI agents and applications, available as an open-source framework and managed platform.
No reviews yet. Be the first to share your experience!
Get started with Decision Node and see if it's the right fit for your needs.
Get Started →Take our 60-second quiz to get personalized tool recommendations
Find Your Perfect AI Stack →Explore 20 ready-to-deploy AI agent templates for sales, support, dev, research, and operations.
Browse Agent Templates →I ran a six-week evaluation of 15 AI developer tools across four workloads: a Flask-to-FastAPI refactor, a Next.js dashboard built from scratch, 23 bug-fix issues sampled from SWE-bench Verified, and a Rust-to-Go port of a 600-line CLI. **Rankings reflect what shipped working cod
Hidden gems in the AI agent tooling space — from browser infrastructure to memory platforms to observability tools. These production-ready tools solve real problems that most developers haven't discovered yet.