Atomic by design
One document can contain many facts. One fact can depend on many sources. Context objects reflect what information means—not where it lives.
What if every AI assistant had instant access to the knowledge your team already knows? ShanText is exploring a new layer of shared, composable context—built for professionals who can't afford to hunt for answers.
Early-stage concept · Professional services · AI infrastructure
A reusable knowledge object—not another folder of files.
Key dates, obligations, events, terms and relationships
Cross-source history, decisions and relevant timelines
Relevant knowledge delivered when it's needed, with a path back to the underlying evidence.
Files are where knowledge gets stored. Context is what people actually need. ShanText explores turning scattered information into living, reusable context objects that AI systems can understand and share.
One document can contain many facts. One fact can depend on many sources. Context objects reflect what information means—not where it lives.
Combine objects into client, matter, engagement, or organization views. Reuse the same context across teams and workflows.
Designed as an independent layer that assistants and professional AI applications could access through common interfaces such as APIs and MCP.
As sources change, derived facts and summaries can be refreshed, reconciled, and versioned—without rebuilding knowledge by hand.
Know where information came from, when it was true, and what supports it. Answers should be traceable, not merely plausible.
Context sharing should respect client boundaries, matter restrictions, and professional confidentiality at every level.
Bring together relevant information from email, documents, systems of record, conversations, and existing context objects.
Extract facts, preserve relationships, reconcile changes, and assemble permissioned views for the people and matters that need them.
Make the right context reachable from an AI assistant during a client call, meeting, or time-sensitive professional workflow.
Imagine using ShanText-backed context in the middle of a conversation. Explore hypothetical examples of the experience we're designing.
Not consistently. The illustrative payment history shows two late payments in the last six months, followed by an on-time payment in October.
Hypothetical context: payment ledger + manager correspondence · Source links available in a real implementationFor law, accounting, and advisory firms, useful context isn't enough. ShanText's envisioned architecture puts source provenance, confidentiality-aware permissions, history, and auditability at the center—not as an afterthought.
We're exploring how context objects can make institutional knowledge reusable across people, teams, and AI systems—at the speed of conversation.
Back to the vision ↑ShanText is an exploratory project, not a generally available product. Security and professional controls described here are design goals.