Sf Documentation Knowledge

Overview

This system programmatically collects all Salesforce documentation from developer.salesforce.com (129 domains, 35,000+ pages), processes it into structured, curated knowledge files, and serves them to LLM agents via:

1. Context Engineering — Pre-compiled Markdown files with _index.md routing tables
2. MCP Server — 12 tools + 4 prompts + 5 resources via Model Context Protocol
3. Knowledge Graph — 53,000+ nodes and 450,000+ edges connecting SF concepts, namespaces, services, and cross-references

No embeddings. No vector stores. No blind chunking.

Quick Start

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "sf-docs": {
      "command": "npx",
      "args": ["-y", "-p", "@sfdxy/sf-documentation-knowledge", "sf-docs-mcp"],
      "env": {
        "SF_ACTIVE_DOMAINS": "apex-guide,apex-reference,lwc"
      }
    }
  }
}

Remove the env block to search all 129 domains. See Domain Restriction for details.

Restart Claude Desktop.

VS Code (GitHub Copilot)

Add to .vscode/mcp.json in your workspace (or globally in VS Code settings):

{
  "servers": {
    "sf-docs": {
      "command": "npx",
      "args": ["-y", "-p", "@sfdxy/sf-documentation-knowledge", "sf-docs-mcp"],
      "env": {
        "SF_ACTIVE_DOMAINS": "apex-guide,apex-reference,lwc"
      }
    }
  }
}

Then use @sf-docs in Copilot Chat to query Salesforce documentation.

Gemini Code Assist / Gemini CLI

Add to your MCP config (~/.gemini/settings.json or project .gemini/settings.json):

{
  "mcpServers": {
    "sf-docs": {
      "command": "npx",
      "args": ["-y", "-p", "@sfdxy/sf-documentation-knowledge", "sf-docs-mcp"],
      "env": {
        "SF_ACTIVE_DOMAINS": "apex-guide,apex-reference,lwc"
      }
    }
  }
}

Cursor

Add in Settings -> MCP Servers -> Add Server:

• Name: sf-docs
• Command: npx -y -p @sfdxy/sf-documentation-knowledge sf-docs-mcp
• Transport: stdio
• Environment: SF_ACTIVE_DOMAINS=apex-guide,apex-reference,lwc

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "sf-docs": {
      "command": "npx",
      "args": ["-y", "-p", "@sfdxy/sf-documentation-knowledge", "sf-docs-mcp"],
      "env": {
        "SF_ACTIVE_DOMAINS": "apex-guide,apex-reference,lwc"
      }
    }
  }
}

OpenCode

Add to your OpenCode config (~/.config/opencode/config.json or project .opencode/config.json):

{
  "mcpServers": {
    "sf-docs": {
      "command": "npx",
      "args": ["-y", "-p", "@sfdxy/sf-documentation-knowledge", "sf-docs-mcp"],
      "env": {
        "SF_ACTIVE_DOMAINS": "apex-guide,apex-reference,lwc"
      }
    }
  }
}

Any MCP Client (Generic)

Point your MCP client to:

npx -y -p @sfdxy/sf-documentation-knowledge sf-docs-mcp

The server uses stdio transport and is compatible with any MCP client.

Why -p + sf-docs-mcp?
The package ships two binaries: sf-knowledge (the data pipeline CLI) and sf-docs-mcp (the MCP server). Using -p installs the package and then explicitly calls the sf-docs-mcp binary, ensuring you get the MCP server and not the CLI.

Use from Source

git clone https://github.com/Avinava/sf-documentation-knowledge.git
cd sf-documentation-knowledge
npm install
npm run build
npm run mcp:start

MCP Server

The MCP server loads the full 53k-node knowledge graph and 18,000+ code snippets into memory on startup (~5s) and serves all queries instantly.

Run directly from source:

npm run mcp:start

Or via npx (no clone required):

npx -y -p @sfdxy/sf-documentation-knowledge sf-docs-mcp

Tools (12)

Prompt Templates (4)

Resources (5)

Agents can read these without making a tool call:

Domain Restriction

When working on a specific Salesforce product area (e.g., Revenue Cloud, Apex development), you can restrict all tools to only search within relevant domains. This reduces noise and improves result quality.

How It Works

• At startup: Set SF_ACTIVE_DOMAINS as a comma-separated list of domain IDs in your MCP client config
• At runtime: Use sf_set_active_domains to change the active domains without restarting
• Not set: All 129 domains are searched (default, no breaking change)
• Per-call domain filter outside active set: Returns empty results with a warning (not an error)
• sf_read_topic outside active set: Shows a gentle note but still allows reading

Discovering Domains

# Let the AI suggest domains for your task
sf_suggest_domains("building LWC components with Apex backend")
→ Suggests: lwc, apex-guide, apex-reference, lightning

# Set the suggested domains
sf_set_active_domains(domains: ["lwc", "apex-guide", "apex-reference", "lightning"])

# Check current state
sf_set_active_domains()

# Clear restrictions
sf_set_active_domains(clear: true)

Behavior by Tool

All 129 Domain IDs

See docs/domains.md for the full list organized by service category, or use sf_list_domains at runtime.

Knowledge Base

The repository comes pre-loaded with 35,000+ curated markdown files and a Knowledge Graph (53,000+ nodes, 450,000+ edges) covering 129 domains of Salesforce documentation.

Option A: Context Engineering (File-based)

Point your AI agent to the _index.md file in any domain folder. The index acts as a routing table telling the AI which files contain which topics:

knowledge/current/<domain-name>/_index.md

Each domain folder also has a SKILL.md in skills/<domain-name>/SKILL.md that teaches AI agents how to navigate the knowledge.

Option B: Knowledge Graph

The graph at knowledge/current/graph.json connects all documentation with semantic relationships:

Inspect it with:

npm run graph:stats

See Graph Schema Documentation for the full schema with node/edge types, ID conventions, and a visual diagram.

Data Pipeline

To update the knowledge base with the latest Salesforce releases, run the pipeline in order:

Step 1: Discover Available Deliverables

npm run discover

Lists all documentation deliverables available from the Salesforce Index API (~127 deliverables).

Step 2: Collect Raw Data

# Collect a specific domain
npm run collect -- --domain cli-commands

# Collect all configured (P0) domains
npm run collect

# Collect ALL deliverables from the SF index API (121 domains, ~31k pages)
npm run collect -- --discover

Step 3: Process HTML to Markdown

# Process a specific domain
npm run process -- --domain cli-commands

# Process ALL collected domains
npm run process -- --discover

Automatically cleans HTML, strips noise, parses tables, formats code blocks, creates clean Markdown, and redacts any Salesforce tokens or secrets.

Step 4: Generate Knowledge Files & Graph

# Generate ALL collected domains and rebuild the full Knowledge Graph
npm run generate -- --discover

Builds the knowledge graph (cross-references, namespaces, service categories, doctype clustering), generates context files, and updates inventory docs.

Step 5: Inspect the Graph

npm run graph:stats

Full Pipeline (One-liner)

npm run collect -- --discover && npm run process -- --discover && npm run generate -- --discover

CLI Reference

All pipeline commands support --domain <name> for single-domain processing and --discover for all-domain processing.

CI/CD

Documentation

License

MIT © Avinava

Inventory

129 domains | 35,438 knowledge files