Use DocInject with AI assistants (MCP)

DocInject exposes your SOPs, playbooks, guides, and handbooks as an MCP server so AI assistants can read, search, and summarize your documents in context. Once connected, you can ask Claude or Cursor questions like “what is our onboarding process?” and get answers grounded in your actual documentation.

Prerequisites

You or an administrator must first create an API key in Profile → API Keys before anyone on the team can connect. See API keys for instructions.

Connect via OAuth (Claude.ai, mobile, and tablet apps)

Apps that support OAuth — including Claude.ai on the web, iOS, and Android — connect using DocInject’s built-in OAuth flow. No config file is needed.

Open your AI assistant and navigate to the MCP or integrations settings.
Add a new MCP server with the URL https://mcp.docinject.com.
When prompted, enter your API key.
The app completes the connection automatically - no browser login or approval screen appears.

Your API key acts as your company wide identity in the context of MCP. The app exchanges it for an access token behind the scenes.

Connect via config file (Claude Desktop, etc.)

Desktop clients that use a config.json file connect by passing the API key directly as a header.

Mac / Linux
Windows

Config file location: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "docinject": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.docinject.com",
        "--header",
        "Authorization: Bearer YOUR_API_KEY"
      ]
    }
  }
}

Config file location: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "docinject": {
      "command": "cmd",
      "args": [
        "/c",
        "npx",
        "mcp-remote",
        "https://mcp.docinject.com",
        "--header",
        "Authorization: Bearer YOUR_API_KEY"
      ]
    }
  }
}

Windows requires the cmd /c wrapper because npx cannot be invoked directly when Node is installed in a path containing spaces (e.g. C:\Program Files\nodejs).

Replace YOUR_API_KEY with the key created by your admin. Restart the app after saving.

Connect directly (advanced)

For clients that support custom HTTP headers natively, point them directly at the MCP endpoint:

URL: https://mcp.docinject.com/mcp
Authorization header: Bearer YOUR_API_KEY

Connect from automation platforms

Tools like n8n and Gumloop support MCP natively and can connect directly to https://mcp.docinject.com/mcp using your API key. Other platforms (Zapier, Make, Pipedream) do not have native MCP support but can call the endpoint using standard HTTP requests.

Platform	Native MCP	How to connect
n8n	Yes	Use the MCP Client node — set the URL to `https://mcp.docinject.com/mcp` and add `Authorization: Bearer YOUR_API_KEY` as a header
Gumloop	Yes	Add a new MCP tool and point it at `https://mcp.docinject.com/mcp` with your API key
Pipedream	Partial	MCP support is available in AI actions; check their docs for the latest setup steps
Zapier	No	Use the HTTP action to send JSON-RPC requests directly (see below)
Make	No	Use the HTTP module to send JSON-RPC requests directly (see below)

Calling tools over HTTP (Zapier, Make, and others)

Platforms without native MCP can call your tools by sending a JSON-RPC request to https://mcp.docinject.com/mcp:

POST https://mcp.docinject.com/mcp
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "list_documents",
    "arguments": {}
  }
}

Replace list_documents with any of the available tools and add the required arguments. Each call is stateless — no session setup is needed.

Available tools

Once connected, your AI assistant has access to the following tools:

Tool	What it does
`list_documents`	Lists published documents in your organization
`search_documents`	Searches published document titles by keyword
`get_document`	Retrieves a document’s full structured content
`get_document_markdown`	Returns a document formatted as Markdown, ideal for summarization
`get_section`	Returns a single section from a document by title, including its steps and sub-steps

The AI chooses which tools to call based on your question. You do not need to invoke them manually.

Injecting sections into other applications

get_section is designed for use cases where you want to pull a specific part of an SOP into another tool — a support ticket, a Notion page, a HubSpot template, or a custom prompt — without bringing in the entire document. You can ask your AI assistant things like:

“Get the ‘Equipment Setup’ section from our onboarding SOP and add it to this email draft”
“Pull the incident escalation steps from DocInject and insert them into this runbook”

The returned section includes a From: Document Title attribution so the source is always clear. Because it reads live from DocInject, the content reflects the current published version of your document every time it is called.

Example prompts

Once connected, you can ask your AI assistant things like:

“Summarize our employee onboarding SOP”
“Find any documents related to security”
“What are the steps in our incident response playbook?”
“List all published guides in DocInject”

Security

The MCP server is read-only. Connected AI assistants can read and search your published documents but cannot create, edit, archive, or delete anything. API keys are scoped to your organization. If a key is compromised, revoke it immediately from Settings → API Keys and generate a new one.

Do not share API keys or commit them to source control. Anyone with a valid key can read your organization’s published documents.

Get Started

Creating Documents

Sharing & Exporting

Integrations

Team & Settings

Use DocInject with AI assistants (MCP)

Prerequisites

Connect via OAuth (Claude.ai, mobile, and tablet apps)

Connect via config file (Claude Desktop, etc.)

Connect directly (advanced)

Connect from automation platforms

Calling tools over HTTP (Zapier, Make, and others)

Available tools

Injecting sections into other applications

Example prompts

Security

​Prerequisites

​Connect via OAuth (Claude.ai, mobile, and tablet apps)

​Connect via config file (Claude Desktop, etc.)

​Connect directly (advanced)

​Connect from automation platforms

​Calling tools over HTTP (Zapier, Make, and others)

​Available tools

​Injecting sections into other applications

​Example prompts

​Security

Prerequisites

Connect via OAuth (Claude.ai, mobile, and tablet apps)

Connect via config file (Claude Desktop, etc.)

Connect directly (advanced)

Connect from automation platforms

Calling tools over HTTP (Zapier, Make, and others)

Available tools

Injecting sections into other applications

Example prompts

Security