← Compare Compare · per product
Hydrate vs AtomicMemory
agent memory infraPluggable, auditable memory engine built around AUDN (Add/Update/Delete/No-op) + contradiction-safe versioning. Convergent with Hydrate's own AUDN unit; Hydrate adds the coding-team adapter + canon + orchestration.
What AtomicMemory is
| Stack / runtime | HTTP service (published Docker image); pluggable embedding/LLM/storage/scope |
| Licence | Open source |
| Storage | Pluggable artifact storage + semantic retrieval |
| Hooks | None (call the HTTP service) |
| Injection | Service API |
| Scope | Standardised platform layer for builders ('not a framework, not SaaS') |
| Maturity | Open source; docs.atomicstrata.ai |
| Price | Free OSS (self-host) |
| Version referenced | (verify) |
| Source | github.com/fcunnane/AtomicMemory |
Where AtomicMemory leads
- Pluggable at every seam (swap embedding/LLM/storage/scope without forking)
- AUDN + contradiction-safe claim versioning as a platform primitive
- Explicit operational-control / auditability thesis
Where Hydrate leads
- Finished coding-team product, not a platform layer (zero-config, automatic)
- AUDN PLUS team canon propagation, cross-vendor live, orchestration
- Single dependency-free binary vs HTTP/Docker service
Verification notes
CONVERGENCE: Hydrate has its own AUDN unit (memory: project_unit_b_audn_build, gated HYDRATE_AUDN). Lead with convergence framing, not feature-gap. Check if arXiv 'AtomMem' paper is the same project before citing.
Sources
- Primary repo https://github.com/fcunnane/AtomicMemory
- Docs https://docs.atomicstrata.ai/