DocForge Docs

1. Go to the releases page and download the installer (.exe or .msi)
2. Run the installer — it takes about a minute
3. DocForge opens automatically in a window when installation completes
4. Go to Settings to connect your AI model before generating your first document

DocForge stores its database at %APPDATA%\DocForge\ on Windows.

Yes. The warning appears because the installer is newly distributed and not yet widely seen by SmartScreen.

1. Click More info on the SmartScreen dialog
2. Click Run anyway

This is a one-time prompt that disappears as more users install.

• Windows 10, 11, or Server 2019+
• ~200 MB disk space
• An AI model API key or Ollama installed locally
• Internet access only if using a cloud AI provider (OpenAI, Anthropic, etc.)

1. Get an API key from platform.openai.com
2. In DocForge, go to Settings → LLM Provider
3. Select OpenAI from the preset list
4. Paste your API key and set the model (e.g. gpt-4o)
5. Click Test Connection then Save

Recommended model: gpt-4o — best quality for long structured documents.

1. Select Azure OpenAI from the preset list
2. Set the Base URL to your deployment endpoint: https://<resource>.openai.azure.com/openai/deployments/<deployment>/
3. Enter your Azure API key
4. Set the model name to match your deployment name (e.g. gpt-4o)

Azure OpenAI uses your existing Microsoft agreement — data stays in your Azure tenant.

1. Get an API key from console.anthropic.com
2. Select Anthropic (Claude) from the preset list
3. Paste your key (starts with sk-ant-)
4. Set model to claude-opus-4-5 or claude-sonnet-4-5

Claude models are excellent for long structured documents with complex exception handling.

1. Get a free API key from console.groq.com
2. Select Groq from the preset list
3. Paste your key (starts with gsk_)
4. Recommended model: llama-3.3-70b-versatile

Groq runs 70B models at ~280 tokens/second — much faster than OpenAI for the same quality. Free tier is generous for most teams.

1. Install Ollama and pull a model: ollama pull llama3.1:8b
2. In DocForge, go to Settings → Models to manage your local models
3. Or select Ollama (Local) in LLM Provider settings manually

Ollama is the best option when data privacy is critical. For high-quality output, use a 13B+ model — 8B models can struggle with long structured documents. A dedicated GPU is recommended.

1. Click New Project on the Projects page and give it a name (the process name)
2. Inside the project, click + Add Input and paste your meeting transcript or process notes — or upload an existing file
3. Select the inputs you want to use and choose a document type (start with PDD)
4. Click Generate — the document streams in live as it's produced
5. Review the document and click Export DOCX to download it

When DocForge can't determine a specific value from your inputs — a name, a volume, a system URL — it inserts a[CLARIFY: your question here] placeholder rather than guessing.

All gaps are collected in the Information Gaps panel after generation. You can answer them directly in DocForge and regenerate, or fill them in manually in Word after export.

Yes. Add new inputs or edit existing ones, then click Generate again. DocForge saves every version — you can compare versions and restore a previous one at any time from the document's version history.

• Pasted text — meeting transcripts, process notes, emails, any plain text
• .docx — existing Word documents (content is extracted as text)
• .pdf — PDFs (text is extracted)
• .xlsx — spreadsheets (data is extracted as structured text)
• .txt / .md — plain text files

Binary content (images, scanned PDFs) is skipped — DocForge only sends readable text to the AI.

By default, DocForge scans your inputs before sending them to the AI and redacts personal information — names, email addresses, phone numbers, and similar identifiers — replacing them with neutral placeholders like [PERSON_1].

You can review exactly what will be redacted before confirming the generation. If a value is intentional (e.g. a named business owner), you can uncheck it to keep the original. Disable the feature entirely in the generation panel if your inputs contain no sensitive data.

When you export a document, DocForge can use your own Word template instead of the built-in style. It opens the template, clears the body content, and fills it with the generated text — preserving every style, header, footer, font, and colour exactly.

The result is a DOCX that looks identical to your organisation's existing document standards.

1. Open one of your existing PDDs (or whichever doc type you're templating) in Word
2. Keep all the formatting — headers, footers, fonts, styles, cover page
3. Delete the body content (the actual text), leaving the structure intact
4. Save as .docx
5. Upload it in DocForge under Templates (global) or inside a project (project-specific)

Pro tip: keep at least one section break so Word doesn't collapse the header/footer on export.

Yes. Open a project and use the Templates panel inside the project to upload a project-specific template. Project templates take priority over global templates for that project.

No. If no template is uploaded, DocForge uses its built-in professional style — clean, Calibri-based formatting that mirrors standard RPA document conventions. You can upload a template at any time and re-export.

• 2 projects
• 5 document generations (lifetime)
• PDD generation only
• DOCX export
• Bring your own AI model (all providers)

No credit card required. Download and use immediately.

• Unlimited projects
• Unlimited document generations
• All document types — SDD, UAT Test Plan, SOP
• Custom DOCX templates (global and per-project)
• Future document types included at no extra cost
• Perpetual license — never expires

1. After purchase, check your email for the license key
2. Open DocForge and go to Settings → License
3. Paste the key (starts with DF-) and click Activate

Activation is instant and works offline — no internet connection needed after activation.

No. DocForge Pro is a perpetual one-time purchase. Your license never expires. There are no subscription fees, no annual renewals, and no usage limits.

Check your spam/junk folder first. If it's not there after a few minutes, email [email protected] with your order reference (visible on the Stripe receipt) and we'll resend it.

This is a known issue with some Windows configurations. Try reinstalling using the latest installer from the releases page. If the error persists, install the Visual C++ Redistributable and try again.

• Go to Settings → LLM Provider and click Test Connection
• Make sure your API key is correct and hasn't expired
• For Azure: confirm the Base URL includes your resource name and deployment name
• For Ollama: make sure Ollama is running (ollama serve) and the model is pulled

This usually happens with smaller local models (under 13B parameters) that run out of context. Try switching to a cloud model like GPT-4o or Groq's Llama 70B for complete output. If you're already using a cloud model, try splitting very long input transcripts into smaller sections.

Make sure you're pasting the full key including the DF- prefix and no extra spaces. If the error says "signature is invalid", the key may have been corrupted in transit — email [email protected] and we'll send a fresh one.

Email [email protected] — include your DocForge version (visible in the sidebar) and a description of what you're seeing.