Aid4Mail 6.3 Launched: Adds App-Only Linked Attachments and Thinking Model Support

We’re pleased to announce the release of Aid4Mail 6.3, an update focused on two things that matter to forensic investigators and eDiscovery teams: collecting more of the evidence that lives outside the message body, and getting more dependable results from AI-assisted review.

This release extends Microsoft 365 linked-attachment collection to App-Only Access, so automated and unattended jobs can now capture cloud-hosted files that were previously reachable only under interactive sign-in. It also adds support for thinking models through Ollama, refreshes the bundled AI model configurations, and makes large batch configurations easier to manage.

Here’s what’s new in Aid4Mail 6.3.

Microsoft 365 Linked Attachments in Unattended Collections

Modern Microsoft 365 messages often replace traditional file attachments with linked attachments—pointers to documents stored in OneDrive or SharePoint rather than in the email itself. Aid4Mail has collected these linked files under delegated (interactive) authentication since version 6.1. Aid4Mail 6.3 extends that capability to App-Only Access.

App-Only Access lets an administrator grant Aid4Mail tenant-wide access to every mailbox without per-user credentials or user interaction—the foundation for centralized, unattended collection across an organization. Until now, that unattended path could not retrieve linked attachments. With 6.3, scheduled and automated jobs collect the linked files alongside the messages that reference them, closing a gap that previously required an interactive session to fill.

For investigations and litigation holds that depend on complete evidence, this means a routine overnight collection no longer leaves cloud-hosted documents behind. For more on how Aid4Mail captures these files and their metadata, see Cloud Attachments.

Thinking Model Support via Ollama

Aid4Mail’s AI features—Filter, Classify, and Analyze—now support thinking models deployed through Ollama, including Gemma 4, Qwen 3.6, Magistral 24B, and Nemotron 3.

Thinking models perform additional internal reasoning before producing a response. On complex classification and analysis tasks—the kind where context, nuance, or multi-step judgment matters—that extra reasoning can improve accuracy over a single-pass answer. Because these models run locally through Ollama, the reasoning happens entirely offline, keeping confidential email content within your own environment. See AI-powered email analysis for how Aid4Mail puts these models to work.

Alongside thinking-model support, the existing Ollama model configurations have been fine-tuned to improve output quality and consistency across filtering, classification, and analysis.

New AI Model Configurations

Aid4Mail 6.3 adds and updates ready-to-use configurations for the latest frontier and open-weight models, so you can select them without editing configuration files by hand:

• Gemini 3.1 Flash-Lite (Google)—now generally available, replacing the earlier preview configuration
• Claude Opus 4.7 (Anthropic)
• GPT-5.5 (OpenAI)
• Grok 4.2 and Grok 4.3 (xAI)
• Qwen 3.6 27B (Dense) and Qwen 3.6 35B (MoE)

These join the configurations already bundled with Aid4Mail, keeping the software current with the latest releases. If you’re weighing which model fits a given matter—balancing accuracy, cost, speed, data residency, and language coverage—our benchmark of 11 AI models is a useful starting point, and the full AI Model Benchmark report includes a decision matrix that maps common constraints to a recommended model.

More Robust AI API Error Handling

AI providers occasionally return transient or recoverable errors during long runs. Aid4Mail 6.3 handles these responses more gracefully, reducing the likelihood that a single provider hiccup interrupts a large classification or analysis job. For collections that process hundreds of thousands of emails through a cloud model, that resilience translates directly into fewer restarts.

Session List Improvements for Large Batches

Two changes make large batch configurations easier to manage:

• Session checkboxes—An optional display mode adds a checkbox to each session, letting you select a subset to include in a batch run. Unchecked sessions are skipped without being removed from the list, so you can run a partial batch and restore the full set later.
• Filtering by state—Sessions can now be filtered by state—editing, running, stopped, and others—directly in the session list, making it easier to locate and act on the sessions you care about in a large configuration.

The interface also gains numerous smaller refinements, including improved high-contrast display support and better keyboard accessibility.

Component Updates

Aid4Mail 6.3 refreshes several bundled third-party components:

• ExifTool updated to version 13.57 for the latest metadata standards
• 7-Zip DLLs updated to version 26.01 for improved archive handling and security
• Aspose.Email for .NET updated to version 26.4

Aid4Mail 6.3 is a focused update that removes a real constraint on automated Microsoft 365 collection and gives AI-assisted review more capable models to work with. Whether you’re running unattended enterprise collections or classifying large datasets offline, this release helps you capture more complete evidence and get more reliable results.

Try Aid4Mail 6.3 today and see what the new collection and AI capabilities can do for your workflow.