The epistemic computing engine for autonomous enterprise systems.
AletheiaGraph combines authoritative graph persistence, enterprise query execution, logical reasoning, evidence provenance, governed traversal and truth semantics in one durable engine.
Why another engine?
If the job were only to store enterprise records, we would use PostgreSQL. AletheiaGraph exists because autonomous systems push correctness and governance requirements below the application boundary and through persistence, query planning, graph traversal, reasoning, proof and authority.
One engine, six roles
Persistent operational state
Transactions, concurrency, recovery, indexes, CRUD, joins, aggregation and pagination.
Relationships are first-class
Vertices, edges and graph-native relationships are part of the authoritative state, not an application-side projection.
Semantics-preserving optimization
Native algebra, statistics, cost planning and multiple access paths with authoritative-result equivalence.
Bounded logical inference
Recursive Prolog reasoning over the graph with budgets, cycle protection, timeout control and query isolation.
Evidence that survives
Conclusions can retain durable relationships to the exact committed vertex or edge versions that support them.
Inference is not authority
Stored, inferred and proposed state remain distinct; authority transition and traversal are explicit rather than implicit.
These are not six attached products. They operate over the same authoritative graph.
The epistemic model
AletheiaGraph treats epistemic state as part of the engine. It preserves the difference between what was stored, what was inferred, what was proposed and what was explicitly promoted into authoritative state.
Explore proof, reasoning and epistemic state →
Reachability is not authority
A conventional graph engine asks whether one node can reach another. An autonomous enterprise system must also ask whether a principal may cross that relationship, under the relevant authority, for the current operation. AletheiaGraph is being designed so graph exploration can be constrained by governed surfaces rather than treating all topology as equally traversable.
Why CRC matters to AletheiaGraph
AletheiaGraph is deliberately being built to do more than record retrieval. It can represent first-class relationships, execute graph-native queries, perform bounded logical inference, preserve proof topology, and ultimately coordinate queries across governed graph surfaces. Those capabilities create power that a passive record store does not have: the engine can discover and exercise paths through organizational knowledge and authority.
CRC supplies the constraint model. Isolation surfaces define local boundaries; traversal envelopes constrain depth, edge classes, reachable nodes and consequence-weighted exposure; federation envelopes constrain cross-surface participation, exchanges, data movement and remote-state uncertainty. AletheiaGraph's optimizer is therefore intended to choose the cheapest permissible plan, not merely the cheapest plan.
First: admissibility
Identity, authority, CRC edge classes, reachability and federation limits remove impermissible plans.
Then: exposure
Among permitted plans, prefer plans that reduce unnecessary boundary crossings, graph expansion and transferred material.
Then: execution cost
Indexes, statistics, ordering, joins and other physical choices optimize execution without changing the permitted graph.
This also creates a natural execution record: what CRC envelope was permitted, what plan was selected, what traversal/federation exposure was actually exercised, what evidence supported the result, and what authority allowed the operation.
Correctness before optimization
AletheiaGraph is engineered for semantic correctness, deterministic state transitions and evidentiary transparency. Performance matters, but optimization is constrained by meaning.
Indexes do not become truth
Indexed results are qualified against authoritative graph semantics.
Plans cannot change meaning
Cost-based execution may choose a different physical path, not a different answer.
Inference is attributable
Derived conclusions remain distinguishable from stored facts and can retain exact supporting evidence.
Maturity is evidence, not elapsed time
Field age matters because production creates unpredictable combinations no finite test program can guarantee it has anticipated. But age is a proxy for accumulated evidence, not the evidence itself. AletheiaGraph deliberately attacks the failure modes most important to an authoritative agent-first graph and preserves discovered failures as repeatable qualification and regression cases.
See the qualification testbed →
Why now?
We likely would have built above PostgreSQL.
The cost of owning durable storage, indexes, planning, recovery and systematic qualification would have made accepting an impedance mismatch the rational choice.
AI changed the economics of scrutiny.
AI-assisted engineering makes continuous implementation review, adversarial test generation, edge-case discovery, comparative design analysis and repeated qualification practical at a level that previously required a much larger specialist team.
One authoritative substrate from transaction to inference
Agent-first requirements created the need. Epistemic computing is the architecture that emerged.
AletheiaGraph can perform the role of an enterprise database, but its purpose extends beyond database management: to provide an authoritative substrate on which autonomous systems can know, reason, explain and act.
What is an epistemic computing engine? →