Cortex memory and project intelligence

What Cortex remembers

Cortex records durable project knowledge so AI workers do not rely on fragile chat history. It keeps identity, decisions, lessons, handoffs, tasks, artifacts, vector recall, code graph context, and knowledge-graph relationships in one governed memory surface.

Why it matters

A long-running AI project fails when the team forgets what was decided. Cortex gives each worker a fresh, project-scoped briefing before work starts, then lets the worker search prior evidence instead of guessing.

• Fewer repeated explanations.
• Clear audit trails for decisions.
• Safer handoff continuity across workers and sessions.
• Better recall for code, docs, and product context.

The user-facing model

Users do not need to understand the storage behind Cortex. The practical model is simple: Cortex remembers project records, connects related knowledge, keeps artifacts attached to decisions, and gives the next worker enough context to continue safely.

The six memory views

Cortex can be explained as six user-facing memory views: project records, semantic recall, project/code map, knowledge graph, artifact memory, and boot briefing. Together they help the system remember what happened, what it means, what it touches, what it relates to, what files matter, and what the next worker needs to know.

• Project records: decisions, lessons, handoffs, tasks, and status notes.
• Semantic recall: search that finds related prior context even when wording differs.
• Project map: a view of affected product areas, files, or components.
• Knowledge graph: connected business concepts, requirements, people, systems, and decisions.
• Artifact memory: important files, diagrams, screenshots, and generated outputs.
• Boot briefing: a concise current briefing before work resumes.

Communication streaming

Cortex also supports the communication story. AI workers can pass work, record why a decision was made, preserve evidence, and let the next role continue without relying on a fragile chat transcript.

What stays private

Project memory is scoped to the project boundary. Public docs describe the system shape; they do not expose customer code, local paths, private handoff IDs, secrets, or internal operational payloads.

Read next

Read Cortex memory next for a deeper user-level explanation of the memory views. Read PROMI next to understand how that memory becomes routed work.