Skip to content

Latest commit

 

History

History
526 lines (411 loc) · 15.7 KB

File metadata and controls

526 lines (411 loc) · 15.7 KB

Tool Integrations & Direct Connections

Last Updated: March 9, 2026
Status: Active ecosystem - review monthly for new capabilities

This guide documents the direct integrations between core HeliOS-Studio tools that eliminate manual handoffs and enable true AI-powered automation.

Overview

The ecosystem has matured significantly with MCP (Model Context Protocol) servers and native API integrations that allow tools to communicate directly:

flowchart TB
  subgraph Discovery
    P[Perplexity] -->|API via n8n| COLLECT[Collection Engine]
  end

  subgraph Knowledge
    COLLECT --> NB[NotebookLM]
    NB <-->|MCP| CLAUDE_KB[Claude Desktop]
    NB -->|Apify export| N8N[n8n Auto-Export]
    N8N --> GH_KB[GitHub Knowledge Repos]
    OBS[Obsidian Vault] <-->|MCP| CLAUDE_KB
  end

  subgraph Reasoning
    CLAUDE_KB[Claude Desktop + MCP]
    CC[Claude Code]
  end

  subgraph Orchestration
    CLAUDE_KB <-->|n8n MCP| N8N
    N8N <-->|webhooks| GH[GitHub]
    N8N <-->|webhooks| CUR[Cursor]
    CC <-->|MCP| GH
  end

  subgraph DevEnv
    CUR <-->|auto via webhooks| N8N
    CUR <-->|GitHub integration| GH
  end

  subgraph Automation
    N8N -->|triggers| EXPERIMENTS[Experiment Runners]
    N8N -->|scheduled| RESEARCH_DIGEST[Research Digests]
  end
Loading

Critical MCP Servers

1. n8n MCP Server

Enables: Claude can build, validate, and deploy n8n workflows conversationally

Installation:

npm install -g @n8n-io/mcp-server

Claude Desktop config (~/.config/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "n8n": {
      "command": "npx",
      "args": ["-y", "@n8n-io/mcp-server"],
      "env": {
        "N8N_API_KEY": "your-n8n-api-key",
        "N8N_BASE_URL": "http://localhost:5678"
      }
    }
  }
}

Capabilities:

  • List all workflows
  • Get workflow details and execution history
  • Create/update workflows from natural language
  • Validate workflow logic before deployment
  • Trigger workflows with parameters
  • Access 525+ n8n node templates

Use case: "Claude, build me an n8n workflow that monitors GitHub issues labeled 'experiment' and triggers my prototype's /test endpoint every hour"

Sources: czlonkowski/n8n-mcp, n8n community


2. NotebookLM MCP Server (Unofficial)

Enables: Claude can query, search, and manage NotebookLM notebooks directly

Installation (choose one):

Option A: roomi-fields variant (basic):

npm install -g notebooklm-mcp

Option B: Pantheon-Security variant (enhanced security):

git clone https://github.com/Pantheon-Security/notebooklm-mcp-secure
cd notebooklm-mcp-secure
npm install
npm run build

Claude Desktop config:

{
  "mcpServers": {
    "notebooklm": {
      "command": "node",
      "args": ["/path/to/notebooklm-mcp/dist/index.js"],
      "env": {
        "NOTEBOOKLM_AUTH_TOKEN": "your-google-auth-token"
      }
    }
  }
}

Capabilities:

  • List all notebooks
  • Search across notebook content
  • Query specific notebooks for information
  • Get source citations
  • Add new sources to notebooks (via unofficial API)
  • Export notebook summaries

Use case: "Claude, search my NotebookLM 'MCP Security Research' notebook for vulnerabilities related to token passthrough and summarize findings"

Sources: mcpservers.org/notebooklm, Pantheon-Security/notebooklm-mcp-secure

⚠️ Note: NotebookLM official API is not yet public (as of March 2026). These servers use browser automation or unofficial endpoints.


3. Obsidian MCP Server

Enables: Claude can read/write your local markdown knowledge vault

Installation:

npm install -g @smithery/mcp-obsidian

Claude Desktop config:

{
  "mcpServers": {
    "obsidian": {
      "command": "npx",
      "args": ["-y", "@smithery/mcp-obsidian"],
      "env": {
        "OBSIDIAN_VAULT_PATH": "/home/user/obsidian-vault"
      }
    }
  }
}

Capabilities:

  • Search notes by title or content
  • Read full note content
  • Create new notes
  • Update existing notes
  • List notes in folders
  • Follow wiki-links and backlinks

Use case: "Claude, check my Obsidian vault for any notes on 'threat modeling MCP servers' and create a new note summarizing best practices from HeliOS-Studio research"

Sources: smithery-ai/mcp-obsidian, Obsidian forum


4. GitHub MCP Server (Official)

Enables: Claude can manage repos, issues, PRs, and run queries

Installation:

npm install -g @modelcontextprotocol/server-github

Claude Desktop config:

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_your_token_here"
      }
    }
  }
}

Capabilities:

  • Search repositories, issues, PRs
  • Create/update issues
  • Create branches and PRs
  • Commit files
  • Manage labels and milestones
  • Get commit history
  • Read file contents

Use case: "Claude, create a new repo from my heliOS-service-template, add initial docs from this conversation, and open 3 issues for Phase 1 setup"

Sources: PulseMCP server directory, Claude Code MCP list


n8n Native Integrations

Perplexity API Integration

Enables: Scheduled or triggered web research queries with structured results

Setup in n8n:

  1. Add Perplexity credentials (API key from perplexity.ai/settings)
  2. Use HTTP Request node with endpoint https://api.perplexity.ai/chat/completions
  3. Configure query parameters:
    • model: "llama-3.1-sonar-small-128k-online" (fast) or "large" (thorough)
    • messages: Your research query
    • return_citations: true

Example workflow:

[Schedule Trigger: Daily 9am]
  ↓
[Perplexity API: Query "MCP security vulnerabilities 2026"]
  ↓
[Parse JSON: Extract citations and summary]
  ↓
[HTTP Request: Create NotebookLM notebook via Apify]
  ↓
[GitHub: Commit summary to kb-opportunities-mcp repo]
  ↓
[Slack: Post digest to #research channel]

Sources: n8n Perplexity integration, YouTube tutorial


NotebookLM Export (via Apify)

Enables: Export notebooks, conversations, and sources automatically

Setup in n8n:

  1. Create free Apify account
  2. Add Apify credentials to n8n
  3. Use Apify node with actor clearpath/notebooklm-api

Apify actor capabilities:

  • Export notebook summaries (markdown/JSON)
  • Export all sources with citations
  • Export audio/video overviews
  • Export mind maps and timelines
  • Trigger Deep Research mode

Example workflow:

[Webhook: New research query]
  ↓
[Perplexity: Get top sources]
  ↓
[Apify NotebookLM: Create notebook with sources]
  ↓
[Wait 5 mins for processing]
  ↓
[Apify NotebookLM: Export summary]
  ↓
[GitHub: Commit to kb-opportunities repo]

Sources: Apify NotebookLM actor, n8n + NotebookLM guide


Claude API Integration

Enables: Chain Claude reasoning into multi-step workflows

Setup in n8n:

  1. Add Anthropic credentials (API key)
  2. Use HTTP Request node or Claude integration node (if available in your n8n version)
  3. Configure model and prompts

Example workflow:

[GitHub Webhook: New issue labeled "opportunity"]
  ↓
[GitHub: Get issue body]
  ↓
[Claude API: "Analyze this opportunity and produce 3 architecture options"]
  ↓
[Parse response]
  ↓
[GitHub: Create new repo from template]
  ↓
[GitHub: Commit architecture docs]
  ↓
[GitHub: Create issues for each option]

⚠️ Limitation: Claude API cannot access MCP servers directly. For MCP-heavy workflows, use Claude Desktop manually or trigger via webhook.

Sources: n8n Claude integration


GitHub Webhooks

Enables: React to all GitHub events (issues, PRs, commits, releases)

Setup in n8n:

  1. Create Webhook node in n8n and copy URL
  2. In GitHub repo → Settings → Webhooks → Add webhook
  3. Configure events: Issues, Pull Requests, Push, Release

Example automations:

  • New issue labeled "experiment" → trigger prototype test suite
  • PR merged to main → build Docker image, deploy to staging
  • Release published → update documentation site, post to social
  • Issue closed → update GitHub Project status field

Sources: GitHub webhook docs, n8n GitHub automation guide


Cursor Webhooks

Enables: Trigger actions when Cursor AI agent completes tasks

Setup:

  1. In Cursor settings → Agent → Webhooks
  2. Add webhook URL (your n8n endpoint)
  3. Select events: Agent completed, Agent failed, Agent blocked

Example workflow:

[Cursor Webhook: Agent completed]
  ↓
[Parse payload: Get changed files]
  ↓
[Run linters and security scans]
  ↓
[GitHub: Create PR if quality checks pass]
  ↓
[Slack: Notify team]

⚠️ Limitation: Cursor webhooks are outbound-only. You cannot trigger Cursor agents remotely yet (as of March 2026).

Workaround: Use GitHub Actions or n8n to commit changes that Cursor watches; configure Cursor to auto-run on specific file changes.

Sources: Cursor webhook docs, Cursor release notes March 2026


GitHub Projects Integration

Recommended for HeliOS-Studio: Use GitHub Projects v2 as your central work management system.

Why GitHub Projects

Native GitHub integration – no sync overhead
MCP and API support – Claude can update via GitHub MCP
GitHub Actions automation – update fields on events
Free and unlimited for personal repos
Flexible views – Kanban, table, timeline, roadmap

Setup

  1. Create project at github.com/users/zebadee2kk/projects

  2. Link all HeliOS-related repos

  3. Add custom fields:

    • AI Agent (select: Claude, Copilot, Cursor, n8n, Manual)
    • Research Source (text: NotebookLM notebook link)
    • Cost Estimate (number: estimated $ spend)
    • Phase (select: Discovery, Research, Architecture, Prototype, Deploy)
    • Health (select: Healthy, At Risk, Dead)
  4. Create views:

    • Opportunities Board: Status columns (Backlog → Validated → Planned)
    • Active Sprints: Filter Status: In Progress
    • Research Timeline: Roadmap view by Phase
    • AI Agent Status: Table grouped by AI Agent field

Automation via n8n

Workflow 1: New research → create project item

[GitHub Webhook: New issue in kb-opportunities-* repo]
  ↓
[GitHub API: Create item in "Opportunities Board" view]
  ↓
[Set fields: Phase=Discovery, Health=Healthy]

Workflow 2: Cursor agent completes → update status

[Cursor Webhook: Agent completed]
  ↓
[GitHub API: Find related project item by repo]
  ↓
[Update field: Status="Ready for Review"]

Workflow 3: Weekly lab report → update health

[Schedule: Friday 5pm]
  ↓
[Claude API: Analyze experiment results from NotebookLM]
  ↓
[Parse recommendations: keep/kill/pivot]
  ↓
[GitHub API: Update Health field for each project]
  ↓
[GitHub: Post summary comment on project item]

Sources: GitHub Agentic Workflows guide, AI agents for GitHub automation


Complete Integration Matrix

From → To Method Status Use Case
Claude → n8n MCP server ✅ Ready Build workflows conversationally
Claude → NotebookLM MCP server (unofficial) ⚠️ Unofficial Query research notebooks
Claude → Obsidian MCP server ✅ Ready Read/write markdown vault
Claude → GitHub MCP server (official) ✅ Ready Manage repos, issues, PRs
n8n → Perplexity API ✅ Ready Scheduled research queries
n8n → NotebookLM Apify actor ✅ Ready Export notebooks automatically
n8n → Claude API ✅ Ready Chain reasoning steps
n8n → GitHub Webhooks + API ✅ Ready React to events, update data
Cursor → n8n Webhooks (outbound) ✅ Ready Trigger on agent completion
GitHub → n8n Webhooks ✅ Ready React to all GitHub events
Claude API → MCP N/A ❌ Not supported Limitation: API can't use MCP
Remote → Cursor N/A ❌ Not supported Limitation: can't trigger agents

Maintenance Strategy

Monthly Ecosystem Review

Given the rapid pace of AI tooling evolution (March 2026), schedule a monthly check:

Review checklist:

  1. Check PulseMCP.com/servers for new MCP servers (8,590+ as of March 2026)
  2. Review n8n integrations page for new native nodes
  3. Check Cursor, Claude, and NotebookLM release notes
  4. Test existing MCP servers still work (APIs change)
  5. Update this document with new capabilities
  6. Update LAST_REVIEWED date at top of file

n8n workflow to automate this:

[Schedule: First Monday of month, 9am]
  ↓
[HTTP: Scrape PulseMCP new servers RSS]
  ↓
[Filter: Only MCP servers relevant to HeliOS]
  ↓
[GitHub: Create issue "Monthly Tool Review - [Month]"]
  ↓
[Add checklist to issue body]
  ↓
[Assign to self]

Version Pinning

For production stability:

  • Pin MCP server versions in package.json
  • Test updates in staging before updating MCP config
  • Keep backup MCP config for rollback: claude_desktop_config.json.backup

Breaking Change Alerts

Subscribe to:


Next Steps

  1. ✅ Document integrations (this file)
  2. ⬜ Install priority MCP servers (n8n, GitHub, Obsidian)
  3. ⬜ Build core n8n workflows (research pipeline, GitHub automation)
  4. ⬜ Set up GitHub Projects with custom fields
  5. ⬜ Test end-to-end flow: Perplexity → NotebookLM → Claude → GitHub
  6. ⬜ Document first successful "opportunity → prototype" run
  7. ⬜ Schedule monthly ecosystem review

Resources

MCP Directories

Integration Guides

Release Notes