What is Learning Nodes?

How do I use Learning Nodes?

FAQs

Build

Marketplace

Elen AI (Phase 2)Coming Soon

MarketplaceMCP Tools

MCP ToolsThe Model Context Protocol (MCP) is an open standard that lets AI agents — in tools like Cursor, Claude, Windsurf, and others — connect to external data and decision systems.The Workstation handles the human-facing decision workflow. MCP Tools handle the programmatic layer — letting AI agents participate directly in the decision network from your IDE or AI assistant, without leaving the context of your work.This is what it means in practice: instead of manually creating a decision record after writing code, your agent does it. Instead of searching for relevant past decisions by keyword, your agent retrieves them by meaning.Elen MCPThe Elen MCP connects any MCP-compatible AI agent to the Elen decision graph — the same graph that powers the Workstation. Agents can query existing decisions, commit new ones, expand decision details, and supersede outdated records.What Elen MCP enables

suggest — before acting, check if a prior decision already covers the problem

commit — record a new decision with domain, constraints, and rationale

expand — retrieve the full context of a decision record by ID

supersede — formally replace an outdated decision with a revised one

log_decision — create a validated decision record with evidence and confidence

search_precedents — find past decisions by domain and minimum confidence

get_competency — view the agent's domain expertise profile

InstallationAdd this to your IDE or agent's MCP config (Cursor, Claude Desktop, Windsurf, or any MCP-compatible client):

{ "mcpServers": { "elen": { "command": "npx", "args": ["-y", "@learningnodes/elen-mcp@0.1.6"] } } }

With Workstation dashboard — to connect decisions to the visual dashboard at app.elen.learningnodes.com, add the ELEN_LOCAL_API env var:

{ "mcpServers": { "elen": { "command": "npx", "args": ["-y", "@learningnodes/elen-mcp@0.1.6"], "env": { "ELEN_LOCAL_API": "true" } } } }

When to use each tool

Before taking action — call suggest to see if a prior decision already resolves the problem. Avoids re-litigating decisions that were already made.

When making a structural choice — call commit with a domain, question, decision text, and constraints. This makes the reasoning visible and searchable.

When fixing a bug or changing course — if the fix invalidates a previous decision, call supersede to replace it rather than leaving a stale record.

When ambiguity arises — call expand on a decision ID to retrieve its full constraints and context.

The goal is visible reasoning — eliminating tribal knowledge and undocumented assumptions from your engineering and product workflows.VectorMCPVectorMCP is a semantic middleware layer for AI agents. It discovers tools and skills from local files, git repos, and live MCP endpoints — then serves them over MCP using natural language search rather than exact name matching.When you have a large number of tools or skills available to an agent, the standard MCP approach requires the agent to know exact tool names. VectorMCP solves this with a local FAISS embedding pipeline: the agent queries by intent, and VectorMCP returns the most semantically relevant tools — with 20–40% better recall on ambiguous queries.Key capabilities

Semantic search — find tools by intent, not name. Local embeddings, no external API keys required.

MCP proxy — run as stdio or SSE transport for Claude Desktop, Cursor, and other MCP clients.

Skill registry — discover, install, and manage community Claude Skills by name or git URL.

Dashboard — built-in UI at GET /dashboard to manage sources, browse tool inventory, and inspect query analytics.

Zero config — npx vectormcp scan ./my-tools discovers, indexes, and serves tools instantly.

Quick startPath 1 — Zero-config scan (fastest):

npx vectormcp scan ./my-tools

Optional live reload: npx vectormcp scan ./my-tools --watch

Path 2 — Claude Desktop / Cursor integration (MCP stdio):

Add to your MCP config: { "mcpServers": { "vectormcp": { "command": "npx", "args": ["vectormcp", "proxy", "--transport", "stdio"] } } }

Then run: npx vectormcp proxy --transport stdio

Path 3 — Full setup with config:

npx vectormcp init

npx vectormcp add-file ./tools.md — add a local tool definition file

npx vectormcp add-repo https://github.com/your-org/tools-repo — add a git repo

npx vectormcp add-skill ./skills/my-skill — add a SKILL.md-formatted skill

npx vectormcp start --watch

Using VectorMCP with Learning Nodes skillsThe Learning Nodes skill library uses SKILL.md files that VectorMCP natively understands. This means you can point VectorMCP at your local .agent/skills/ folder and get semantic search over all installed skills — including the Elen MCP skill, the marketplace wiki skill, and any custom skills your team has created.

npx vectormcp scan .agent/skills --watch

The agent can then query by intent — "which skill helps me commit a decision to the graph?" — rather than needing to know the exact skill name.Submit a ToolBuilding an MCP tool that extends the decision workflow? The MCP Tools section is designed to grow. Tools that plug into the ASK → DECIDE → ACTIVATE lifecycle — or that extend the Elen decision graph — are a natural fit.Reach out on Discord to discuss integration patterns, or open a PR to the VectorMCP registry to publish your skill.

PreviousData MarketplaceNextIdentity Solutions