AllegroGraph 9.1 Introduces GraphTalker 2.0: Building Persistent Enterprise Memory for AI Agents

New Context Management capabilities help organizations preserve domain expertise, incorporate human feedback, and bring documents and images into Knowledge Graph analysis.

AI agents are becoming increasingly capable of answering questions and working with enterprise data. But producing a useful answer is only part of the challenge. Enterprise AI also needs to understand what data means, how information is connected, and which knowledge and conventions should inform its analysis. Just as important, organizations need ways to preserve that context and improve it over time.

With the release of AllegroGraph 9.1, Franz introduces GraphTalker 2.0, a major update to its agentic natural-language interface for Knowledge Graphs. The release advances Franz’s vision of the Context Graph as a persistent enterprise memory layer for AI agents, with new capabilities for managing context, preserving reusable analytical work, and incorporating documents and images into graph-based analysis.

From Knowledge Graphs to Persistent Enterprise Memory

A Knowledge Graph establishes a foundation of entities, relationships, rules, provenance, and historical information. For AI agents, that connected knowledge provides something a language model alone cannot: enterprise-specific meaning and an inspectable way to test answers against underlying data.

A Context Graph builds on that foundation by connecting knowledge with operational context—such as states, actions, goals, decisions, and supporting evidence—to help agents understand not simply what is known, but why it matters in a particular situation.

GraphTalker brings natural-language interaction to that semantic foundation. Instead of generating a one-shot query, it examines schemas, considers query examples, runs SPARQL queries, interprets results and errors, and refines its approach. GraphTalker 2.0 adds an important capability: a managed way to preserve and maintain the domain guidance that informs this process.

Introducing Enterprise Context Management

The centerpiece of GraphTalker 2.0 is its new Context Management System. Organizations can maintain repository-specific context files containing ontology explanations, business terminology, domain conventions, and proven query patterns. This means a domain expert’s guidance does not have to remain buried in an individual prompt or be re-explained every time a new analytical session begins.

The new browser-based Context workspace allows authorized users to create, upload, edit, and manage those files. Built-in version history makes it possible to compare changes, review previous versions, and restore earlier content. Entire context-file collections, including their histories, can be exported and imported to support reuse, migration, and backup.

These capabilities also create a practical path for human experts to contribute feedback, corrections, and insights to the knowledge guiding future interactions. Rather than treating every response as automatically learned truth, teams can review what worked, curate the appropriate guidance, and maintain a traceable record of changes. The result is enterprise memory that people can inspect, refine, and reuse.

Analyze Documents and Images Alongside the Graph

Real-world questions often involve more than records already in a Knowledge Graph. A researcher may want to discuss a PDF, an analyst may have a CSV export, or an operations team may want to reference a screenshot while investigating connected data.

GraphTalker 2.0 supports attaching PDFs, images, plain-text files, and CSV files to natural-language requests, bringing outside material into the same analytical interaction as AllegroGraph data. Attachments add information to a request; they are not automatically ingested into the Knowledge Graph or turned into permanent enterprise memory.

A More Productive Workspace for Knowledge Exploration

GraphTalker 2.0 also introduces a completely redesigned interface organized around four areas: Chat, Context, Queries, and History. Users can switch among available repositories, maintain domain guidance, browse saved queries and visualizations, and return to saved sessions.

Behind the scenes, reorganized storage preserves queries, visualizations, and analytical sessions, while preprocessed schema overviews help GraphTalker navigate very large enterprise schemas. These are practical improvements for teams working across complex data models and multiple repositories.

Why This Matters for Trusted Agentic AI

Organizations do not simply need AI that can produce plausible answers. They need AI systems whose analysis is informed by the right enterprise knowledge, whose results can be examined, and whose domain guidance can be maintained by the people responsible for it.

AllegroGraph combines Knowledge Graph technology, vector search, symbolic reasoning, and LLM integration to support this Neuro-Symbolic approach. GraphTalker 2.0 extends it by bringing managed context and reusable human expertise closer to everyday AI analysis.

“AI agents need more than access to data. They need to understand what that data means, how it relates to other information, and the business context in which decisions are made.”

— Dr. Jans Aasman, CEO, Franz Inc.

Franz was identified as a Context Graph sample vendor in Gartner’s Hype Cycle for Agentic AI (April 2026), reinforcing the growing industry attention on context as an essential part of agentic AI architectures.

Explore the new capabilities: Visit the GraphTalker product page or review the GraphTalker 2.0 documentation.

Learn more about Context Graphs: Dr. Aasman is scheduled to present “Context Graphs: The Missing Memory Layer for AI Agents” at ODSC AI on October 27, 2026.

Frequently Asked Questions About GraphTalker

What is GraphTalker?

GraphTalker is AllegroGraph’s agentic natural-language interface for enterprise knowledge graphs. It helps users and applications explore graph data by inspecting schema, generating and executing SPARQL queries, refining its approach, and presenting inspectable results.

Do I need to know SPARQL to use GraphTalker?

No. Users can ask questions in natural language. GraphTalker generates and tests SPARQL queries behind the scenes, while making the queries available for review when needed.

What is new in GraphTalker 2.0?

GraphTalker 2.0, included with AllegroGraph 9.1, introduces browser-managed, version-controlled repository context files, an updated workspace, reorganized storage for saved queries and sessions, schema overviews for large models, and support for document and image attachments.

How does GraphTalker help AI agents use enterprise context?

GraphTalker uses the connected facts and relationships in AllegroGraph together with repository-specific guidance about an organization’s ontology, terminology, and query patterns. This gives AI-assisted analysis a more structured basis than a standalone prompt.

Does GraphTalker automatically learn from every answer?

It is possible. But, a generated answer is not automatically treated as verified organizational knowledge.  GraphTalker can preserve analytical artifacts, and teams can update version-controlled context files with reviewed domain guidance.

Can GraphTalker analyze PDF documents and images?

Yes. GraphTalker 2.0 supports attaching PDFs, images, text, and CSV files to individual requests. These attachments provide context to that interaction but are not automatically added to the AllegroGraph database.

How do AllegroGraph and GraphTalker work together?

AllegroGraph stores and manages enterprise knowledge as a semantic graph, with graph querying and reasoning capabilities. GraphTalker is the agentic natural-language interface that helps users and applications investigate that knowledge and inspect supporting query results.