The language behind cited company memory.

Durable company context stored with source, current version, and permission scope.

A memory can be a decision, policy, support resolution, sales handoff, engineering note, or source-backed fact that your AI should be able to recall later.

The act of asking MemoryCrow for an answer from stored memory.

Recall is not the same as adding memory. MemoryCrow plans meter the memory you add, while cited recall stays unlimited.

An answer that includes the source trail proving where the answer came from.

Citations let a person verify the decision, policy, or source behind the answer instead of trusting an AI assertion.

The linked evidence attached to a memory or answer.

The source trail can point back to a Slack thread, Notion page, GitHub PR, Linear issue, support ticket, or another connected source.

The source MemoryCrow treats as the active version of a fact.

When knowledge changes, the current source should become the answer. Older sources can remain traceable without continuing to drive recall.

An older fact that remains in history but should no longer become the answer.

This is how MemoryCrow keeps useful context without letting stale policies, prices, or rollout notes leak into new answers.

Recall that only returns memory within the requester’s allowed MemoryCrow scope.

MemoryCrow filters by personal, workspace, or group scope before returning facts or citations, so private memory does not become a shared answer by accident.

The access boundary of the person or agent asking a question.

Requester scope means the same memory layer can answer one teammate and abstain for another when the saved memory is not available to both of them.

A memory saved for a specific person, workspace, or group.

Examples include private customer notes, leadership-only plans, deal economics, incident details, and restricted engineering discussions.

A deliberate no-answer result when MemoryCrow cannot prove the answer safely.

Abstention happens when the source is missing, stale, unresolved, or outside the requester scope.

MemoryCrow does not train foundation models on customer workspace data.

Your memory is stored and recalled for your workspace. It is not used to train general-purpose models.

A record of the source and version history behind memory and recall.

Audit trail helps teams understand why an answer was returned and what source or change made it current.

An integration that helps MemoryCrow capture, import, or recall source-backed memory.

Connectors can involve sources like Slack and Notion, planned sources like GitHub and Linear, or recall surfaces like Claude, ChatGPT, Cursor, and VS Code.

Model Context Protocol, a standard way for AI tools to call external tools.

MemoryCrow exposes company memory over MCP so compatible AI clients can recall cited, current, permission-safe context.

An AI app or developer tool that can connect to an MCP server.

Claude, ChatGPT, Cursor, VS Code, and other compatible tools can act as MCP clients for MemoryCrow recall.

Saving a useful fact, decision, or handoff into MemoryCrow from a connected workflow.

Writes can come from a person or an authorized AI agent, and should keep the source and scope attached.

The shared memory layer available to a team workspace.

Workspace memory is useful when multiple people and AI tools need the same current answer with permission checks.

A one-time add-on that increases room for large imports or backfills.

Memory packs are for adding more stored memory. They do not unlock recall because recall is unlimited on every plan.

Data Processing Addendum, the terms that govern how MemoryCrow processes customer data.

Teams use a DPA to understand processing roles, safeguards, subprocessors, and data handling commitments.

A third party MemoryCrow uses to process data while providing the service.

The sub-processor page lists vendors involved in hosting, infrastructure, communication, analytics, or other necessary service operations.

The party that determines why and how personal data is processed.

In many workspace contexts, the customer controls the source data they connect and the purposes for using MemoryCrow.

The party that processes personal data on behalf of a controller.

MemoryCrow may act as a processor for customer workspace data under the relevant agreement.

A database optimized for similarity search over embedded content.

Vector search can be part of retrieval, but it is not the same as memory with source, current version, and permission scope.

Finding relevant content for an AI answer.

Retrieval is useful, but MemoryCrow focuses on the trust layer around retrieval: citations, currency, and access control.

The origin and history of a fact or answer.

Provenance matters because a company answer is only useful when you know where it came from and why it should be trusted.

The amount of information an AI model can consider in one interaction.

A larger context window is not the same as durable company memory. Memory should survive across tools, people, and time.

A customer-facing rule or resolution support needs to answer accurately.

Examples include refund windows, SSO troubleshooting, downgrade handling, and enterprise SLA details.

A commitment made to a customer or prospect that should be remembered with scope.

Customer promises often live in email, CRM notes, calls, and Slack threads. They need citations and permissions.

A technical decision and its reasoning that future engineers or AI assistants may need.

Examples include architecture tradeoffs, rollout decisions, incident lessons, and why a previous approach was abandoned.