Permission-aware context
The design should respect who can read and change stored context, and how that access changes over time.
Planned product direction · Not available today
Useful context beyond a single conversation.
Srasta Membrane is our planned direction for persistent, permission-aware memory: the decisions, constraints, and work state that later tasks may need.
We want to make that context manageable and reviewable. Scope, release timing, and pricing are not yet committed.
Requirements shaping the design
Memory needs more than storage.
Sources and corrections
The design should make it possible to trace where context came from, review a change, and correct outdated information.
Retention and deletion
The design should make ownership, expiry, and deletion explicit, including boundaries between teams and environments.
These are design goals, not currently available Membrane features.
Stored context, not automatic retraining.
Persistent memory would supply relevant context to a model when needed. It would not, by itself, change the model’s weights or guarantee that an answer is correct.
Document retrieval is already a separate platform capability. You can evaluate knowledge workflows today without depending on future Membrane availability.
Explore current knowledge workflows →Part of a broader direction
We also plan an agent harness for companies building their own agents with models, tools, and memory. Its execution and approval requirements are separate from memory storage.
Explore the planned agent harness →Questions about Membrane
Is Srasta Membrane available today?
No. Membrane is a planned product direction. Scope, release timing, and pricing are not committed. Evaluate current Srasta capabilities independently of this roadmap.
What does long-term memory mean here?
The direction is persistent, permission-aware context: decisions, constraints, and work state that can be retrieved for later tasks. It does not mean automatically retraining a model’s weights.
How is this different from document retrieval?
Document retrieval supplies relevant source material for a task. Membrane is intended to manage persistent context across tasks, with requirements for provenance, permissions, correction, and deletion.
Where would memory data live?
The planned direction is customer-controlled storage with explicit access boundaries. The final deployment design is not yet committed. Connected clients, tools, hosted services, and telemetry must each be reviewed as part of the complete data flow.
Help define a useful memory workflow.
Bring an example of context your team needs to retain, who should access it, and when it should be corrected or removed.