Home / Compare / Heartwood Memory

Heartwood proves who could see it. Lians proves when you knew it.

Heartwood centers access policy and signed provenance. Lians centers bitemporal facts and historical reconstruction.

Credit first

Heartwood has a real governance thesis.

Heartwood enforces policy before ranking, signs provenance, isolates tenants, chains its audit log, and follows erasure into derived embeddings.

These are meaningful controls; the products solve different primary problems.

Documented product surfaces, reviewed July 17, 2026

Same risk class, different primary proof.

CapabilityLiansHeartwood Memory
Primary guaranteeKnowledge state at time TPolicy-compliant, attributable recall
Temporal modelValid time plus system timeEffective-time filtering documented
Historical APIrecall_at and reconstructPublic quickstart centers recall and explain_recall
Revision handlingDeterministic keyed supersessionTyped, governed memory records
Policy before rankingBarrier and namespace controlsCore documented design
Provenance and auditTamper-evident audit and lineageSigned provenance and hash-chained audit
Crypto-shred erasureKey destruction with retained certificatePer-subject key-destruction proof
Backtest lookahead checkbacktest_checkNot listed in the public MCP quickstart
Current core licenseApache-2.0BSL 1.1 from 0.2.0; earlier releases remain MIT
Published SDK surfacesPython, TypeScript, Java, C, and GoPython core; governed TypeScript SDK listed as in development

This is a documentation comparison, not a scored benchmark. Heartwood descriptions link to its public documentation; Lians behavior is reproducible from the Apache-2.0 repository.

Decision guide

Choose the proof your system must produce.

Choose Heartwood when access policy and signed provenance are primary.

Choose Lians when historical knowledge state is primary.

For both needs, evaluate them as complementary control planes.

Sources

Verify the claims directly.

Make time part of the retrieval contract.

Start with Lians