Heartwood vs Cognee — provable records vs graph/ontology memory
TL;DR
Cognee combines relational, vector, and graph stores to turn source data into searchable, connected memory, with tenant/dataset-level permissions and documented audit logging; it leads on graph and ontology depth and has a longer public history than Heartwood. Heartwood's difference is record-level cryptographic proof — signed memories, policy-before-ranking, key-destruction erasure — that you re-execute yourself.
At-a-glance table
| Axis | Heartwood Memory | Cognee |
|---|---|---|
| Governance granularity | Per individual memory record | Tenant roles + direct dataset-level read/write/delete grants |
| Provenance signing | Per-record Ed25519, fail-closed read | Not evaluated in public docs (document-level metadata/provenance) |
| Tamper-evident audit | Hash-chained audit log | Not evaluated in public docs (Cloud lists audit logging) |
| Policy-before-ranking | Yes | Not evaluated in public docs (results limited to accessible datasets) |
| Erasure / RTBF | Crypto-shred + key-destruction proof | Item/dataset/all deletion of graph/vector/relational data, with caveats (no cryptographic proof) |
| Tenant isolation | Yes | Yes — complete data isolation; workspace = tenant |
| Interface | Python library + governed MCP server | Python API, HTTP API, CLI, MCP tools |
| Deployment | Self-hosted, embedded | Local/embedded, self-hosted Docker, managed Cloud, Enterprise BYO cloud |
| License | Source-available (BSL 1.1); 0.1.x MIT | Apache-2.0 OSS engine; Cloud + Enterprise separate |
| Pricing | Free / Team $349·mo / Pro $6,000·yr (early access) | Free ($0, 1 workspace/1M tokens) / Standard $2.50 per 1M tokens + $5/workspace / Enterprise custom |
| Best-for | Record-level, re-executable proof | Graph/ontology-centered memory pipelines |
“Not evaluated in public docs” = not found in Cognee's current primary documentation as of 2026-07-15; it is not a claim the feature is absent. Vendor pricing reverified immediately before publish.
Comparison by dimension
Governance
Cognee governs at the dataset level with access controls and telemetry, and it leads on the knowledge-graph and ontology model. Heartwood governs at the record level: per-memory signatures, policy-before-ranking, and crypto-shred erasure. If cognee is further along on graph depth, Heartwood's lane is record-level proof.
Deletion
Cognee deletion removes graph, vector, and relational data (with raw-file/shared-node caveats); Heartwood adds a per-subject key-destruction proof.
Who Cognee is best for
Teams building graph/ontology-centered memory pipelines with OSS or managed deployment.
Who Heartwood is best for
Regulated teams that need per-record, re-executable proof.
Can Heartwood run underneath Cognee?
As the governed store behind shape-compatible memory calls (example contract, not an official integration).
See the governed-memory modelFAQ
Does Cognee have access control?
Yes — tenant roles plus direct dataset-level read/write/delete grants. Heartwood's difference is per-record, re-executable proof.
Is Cognee open source?
Cognee's engine is Apache-2.0; hosted Cloud and Enterprise are separate.