Letta’s agents manage their own memory. Omnia’s memory follows your privacy rules.
In Omnia, redaction, consent and erasure are set by the platform, and the agent can’t edit them. In Letta, from the MemGPT team, the agent decides what goes into its own memory.
Their card and phone numbers were redacted before anything reached long-term memory. Withdrawing consent covers what was stored over those six weeks, and the erasure runs as a data subject request with an audit trail. docs
on letta Background dreaming reviews recent conversations and rewrites memory, and its review step doesn’t ask for approval. Consent, PII redaction and erasure lifecycles aren’t documented. their docs
At a glance
| Topic | Omnia | Letta, per its docs |
|---|---|---|
| Stateful agent runtime | yes Yes. docs | documented Yes. source |
| Redaction before storage | yes Yes. docs enterprise | not documented Not documented. source |
| Consent and erasure | yes Consent revocation and a DSAR lifecycle. docs enterprise | partly Per-item deletes. source |
| Simulated users before release | yes Yes. docs | not documented Not documented. source |
| Stepped rollout, auto rollback | yes Yes. docs | not documented Not documented. source |
| Kubernetes | yes Helm. docs | not documented Not documented. source |
Common questions
- Is Letta open source?
- Its GitHub repository is Apache 2.0. Omnia is proprietary and free to run; PromptKit and PromptArena underneath are Apache 2.0.
Put your own agent through the gate.
Install Omnia in your cluster and run it free while you build.
Letta claims are from Letta’s published documentation, checked 22 October 2026. Spot something out of date? Tell us.