Getting Started with docrew: Your First Analysis in 5 Minutes

What docrew is

docrew is a desktop AI agent that works with your documents. It runs on your computer, reads files directly from your hard drive, and performs analysis, extraction, comparison, and research tasks through a chat interface. You describe what you need, and the agent does the work -- reading files, processing content, writing outputs -- without you having to upload anything to a cloud service.

The agent is not a chatbot that answers questions about documents you paste into a text box. It has direct access to your file system. It can open a PDF, read a spreadsheet, parse a Word document, write a CSV, and execute multi-step workflows, all from a single conversation. You talk to it like you would talk to a capable colleague: "Pull the key financials from this quarterly report" or "Compare these two contracts and tell me what changed."

This guide walks you through installing docrew, setting up your first workspace, and completing your first document analysis. The whole process takes about five minutes.

Download and install

docrew runs on macOS, Windows, and Linux. Download the installer from docrew.ai and run it.

On macOS, open the downloaded DMG file and drag docrew to your Applications folder. The first time you launch it, macOS will ask you to confirm since it is downloaded from the internet. Click Open.

On Windows, run the installer executable. It installs to your user directory by default. No admin privileges required.

On Linux, download the AppImage or the deb/rpm package for your distribution. The AppImage runs directly after making it executable. The package installs through your system's package manager.

After installation, launch docrew. It opens as a desktop window with a sidebar and a main conversation area. The app also places an icon in your system tray (macOS menu bar, Windows taskbar, or Linux notification area) so it stays accessible even when the main window is closed.

Create an account

When you first launch docrew, you will see a sign-up screen. Create an account with your email address. This account handles authentication, billing, and sync between devices. Your documents never leave your computer -- the account is for the service layer, not for storing your files.

After signing up, you land on the main interface: a sidebar on the left showing your workspaces and conversations, and a large conversation area on the right.

Set up your first workspace

A workspace in docrew is a folder on your computer where the agent operates. It defines the scope of what the agent can access. When you point docrew at a workspace folder, the agent can read and write files within that folder and its subfolders.

To create your first workspace, click the workspace selector in the sidebar and choose a folder. Pick a directory that contains documents you want to work with -- maybe your Documents folder, a project folder, or a specific client folder.

The agent now has access to everything inside that folder. It will not touch files outside of it.

Your first conversation: ask a question about a document

Start a new conversation by clicking the new conversation button. You will see a text input at the bottom of the screen -- this is where you talk to the agent.

Drop a document into the conversation. Drag a file from your file manager -- a PDF, a Word document, a spreadsheet -- and drop it into the chat area. Or simply refer to a file by name if it is in your workspace.

Now ask a question. Here are some examples depending on what you dropped in:

For a PDF report: "Summarize this report and list the three most important findings."

For a contract: "What are the key dates in this agreement -- effective date, expiration, renewal deadlines?"

For a spreadsheet: "What is the total revenue across all rows, and which product category has the highest margin?"

Type your question and press Enter.

Watch the agent work

After you send your message, the agent begins working. You will see its activity in the conversation as it progresses.

First, the agent reads your file. It uses its built-in document parsers to extract the content -- text from PDFs, structured content from DOCX files, data from XLSX spreadsheets. This happens locally on your machine. The file is not uploaded anywhere.

Next, the agent analyzes the content. It sends the extracted text to the language model for reasoning. The model processes your question against the document content and produces an analysis.

Then the agent responds. You see the answer appear in the conversation, streamed in real time as the model generates it. For a straightforward question about a single document, this entire process takes 10 to 30 seconds depending on document length.

The response will be specific to your document. If you asked for key dates in a contract, you will see the effective date, the termination date, and any renewal deadlines, pulled directly from the text. If you asked for a summary of a report, you will get a concise breakdown of the main points.

This is not a template or a generic answer. The agent read your specific file and answered your specific question based on what it found.

Try a second task: extract data into a table

Now try something more structured. Tell the agent: "Read the PDF I just shared and extract all the financial figures mentioned -- item, amount, and page number where each appears. Format the results as a table."

The agent will go back through the document, identify every monetary figure, note what it refers to and where it appears, and present the results in a structured table right in the conversation.

If you want that data in a file you can use elsewhere, follow up with: "Write those results to a CSV file in my workspace." The agent creates the file and saves it directly to your workspace folder. You can open it in Excel, Google Sheets, or any spreadsheet application.

This two-step interaction -- analyze then export -- is a common pattern with docrew. You explore the data conversationally, then output the results in whatever format your downstream workflow needs.

What just happened

In those two interactions, several things happened that are worth understanding.

Local file reading. The agent read your document directly from your hard drive. The file was parsed by docrew's built-in document parsers -- a DOCX parser for Word files, an XLSX parser for spreadsheets, and a PDF text extractor for PDFs. The original file never left your computer.

AI reasoning. The extracted text was sent to a language model for analysis. The model understood your question, identified the relevant information in the document, and formulated an answer. This is more than keyword search -- the model understands context, meaning, and relationships between data points.

Tool use. The agent did not just chat with you. It used tools: a file reading tool to access the document, and a file writing tool to create the CSV. The agent decides which tools to use based on what you ask. You do not need to know about the tools -- you just describe what you need.

Conversation context. The agent remembers the conversation. When you said "write those results to a CSV," the agent knew which results you meant because it had the context from the previous message. You can continue building on the conversation -- refining the analysis, asking follow-up questions, requesting different formats.

Workspace scoping. The CSV file was saved inside your workspace folder. The agent operates within the workspace boundary. It reads from and writes to your workspace, keeping everything organized in one location.

Navigating the interface

The sidebar shows your conversations, organized within folders. You can create folders to group related work -- one for client projects, another for internal documents. Conversations can be favorited or archived.

The conversation area is where the interaction happens. Agent responses stream in real time.

The workspace selector lets you switch between different workspace folders. If you work across multiple projects, each can have its own workspace.

What to try next

Once you are comfortable with single-document analysis, try these more advanced workflows.

Batch processing. Put multiple files in a folder and ask the agent to process all of them: "Read every PDF in the invoices folder and extract the vendor name, invoice number, date, and total amount. Write the results to a single CSV." The agent processes each file and consolidates the results.

Multi-document analysis. Ask questions that span multiple files: "Compare the Q1 and Q2 financial reports and highlight the biggest changes in revenue and expenses." The agent reads both files and produces a comparative analysis.

Iterative refinement. Start with a broad question, then narrow down. Ask for a summary first, then ask about specific sections, then request a structured extraction. The agent maintains context across the conversation, so each question builds on the previous answers.

File creation. Ask the agent to create new documents based on its analysis: "Based on the data in this spreadsheet, write a one-page summary report and save it as a DOCX file." The agent generates the content and writes the file to your workspace.

docrew is designed for document work that is too tedious to do manually and too sensitive to upload to a cloud service. The agent handles the reading, parsing, and analysis. You focus on the decisions that matter.

Start with one document and one question. You will quickly find more uses than you expected.