Skip to content

Local Lore

Local-first, built in stages ​

Lore is pivoting to memory that lives on your devices. The local engine is a developer preview, not a finished replacement for every hosted feature.

Available in the source previewStill being built
Private local projects and durable memoryAutomatic repository and conversation capture
Keyword and local semantic searchTeam identity and encrypted device-to-device sync
Original evidence, revisions, and retirementLocal dashboard and shared task coordination
Project-bound stdio MCP connectionsGuardrails, agent pauses, and human resume
Restart recovery and multiple agents on one deviceInstallers, upgrades, and backup UI

How teams will work ​

Each teammate will keep an authorized local copy of shared memory. Devices will exchange updates when connected. An optional customer-owned always-on machine can hold updates for teammates who are offline. This transport is not implemented yet; sharing a local project today only connects agents on the same computer.

Offline lookups can use the last synchronized state. New team updates require a connection. Access revocation cannot erase a copy from a disconnected device.

Guardrails and accountability ​

The plan is to evaluate cached policies locally, record violations, and pause supported agent integrations when required. Enforcement needs a host that honors those decisions. Lore cannot control arbitrary software on a customer's computer. Cross-team reports reconcile when devices reconnect.

What we host ​

Railway serves the public website and documentation. The local memory runtime does not need a Lore account, hosted database, Redis, GPU, or paid embedding API. Optional billing and license administration can be added when needed. Customer memory and team synchronization will not depend on those services.

The old hosted service is being retired. Its infrastructure and data are retained during the transition; hosted documentation describes historical behavior.