AI context and agent platforms
AI needs a repository of work, not another temporary context window.
RAG and agent platforms help models retrieve information and call tools. Legra addresses the durable corpus those sessions should improve: semantic context, files, conversations, permissions, branches, tasks, evidence, agreements, and operational knowledge organized around graph objects.
Axis coverage
The six axes of the complete logistics loop. What each axis asks.
Compare the mechanisms.
AI context and agent platforms
Fast retrieval of relevant text, vectors, memories, and tool descriptions.
Model routing, agent loops, tool invocation, and conversational interfaces.
Rapid experimentation with changing model capabilities.
Legra
The graph and workspace remain durable and model-provider independent.
Agents receive explicit Use Cases, Stories, authority, budgets, review paths, and provenance.
Human and agent work continuously improves one shared, versioned corpus.
Questions to use in an evaluation
Ask these in the room.
- 01
What durable corpus does each agent session improve?
- 02
Can the system explain authority, inputs, branches, decisions, and resulting changes?
- 03
Will organizational knowledge survive a switch of model or agent vendor?
Graph features are only the substrate.
Legra also organizes the work that uses and changes the graph, the evidence needed to review that work, and the economic relationships that move data and value between parties. That operational layer is what turns graph infrastructure into data logistics infrastructure.
A complete history of the work
Legra keeps each piece of work together with the changes it made, the evidence behind it, and any costs incurred. That history stays private to the workspace and cannot be quietly rewritten.
- Work by people, AI agents, and automated services appears in one traceable record.
- See what is happening now, then return to the same history later for review.
Layered provenance
Full provenance connects signed commits and exact deltas to task and transaction lineage, actors, delegated authority, imports, and every retained rule and premise behind derived facts.
- Evidence answers who or what acted, on whose authority, from which inputs, and with which result.
- Commit, activity, fact-origin, and derivation evidence remains attached to the workspace and its history.
Composable transaction pricing
The admitted total composes independently priced data value, compute, storage, custody, transport, external I/O, rights, and operator services instead of treating bytes as the value of data.
- Charges, credits, refunds, penalties, and commissions contribute to the final amount.
- Zero is an explicit price through the same accounting path, including self-operated work.
Multidirectional token settlement
Every resource-consuming operation follows one path: measure, price, authorize, execute, record, and settle. Detailed evidence stays with the workspace instead of one central billing ledger.
- Tokens can flow to data owners, infrastructure and service Operators, and the NetworkOperator treasury.
- The same evidence supports internal chargeback and external settlement; subscriptions supply token packages rather than feature gates.