Context and memory

A CoAgent has a continuing conversation in its Codex or Claude Code runtime, persistent files, and access to shared knowledge. These serve different purposes.

Conversation context

The runtime resumes the CoAgent's saved conversation across work. Channels notifications, Cortex work, and scheduled tasks can enter that shared context. They are not separate, isolated model sessions for each issue or conversation.

A new event can steer an active turn. The CoAgent must distinguish which conversation or task each instruction belongs to and communicate results in the right place.

Durable memory

Files in /home/agent persist beyond a turn or computer restart. The home template tells the CoAgent to record useful decisions, unfinished work, and reference notes in its memory and documentation folders.

Persistence does not mean every file is included in every model request. The CoAgent must read the relevant material. Long conversations may also be compacted; do not rely on verbatim recall of everything previously said.

Shared knowledge in Cortex

Use Cortex for knowledge other people and CoAgents should retrieve and maintain together. Use local files for working state and notes specific to that CoAgent.

The CoAgent can preserve decisions and corrections from ordinary work in Cortex, maintain their dependencies and retrieve the relevant Modules before acting. You do not need to author a knowledge base first. Give it the required access and guidance, and choose whether it saves changes or recommends them for review.

For an ongoing responsibility, the CoAgent finds the governing Modules and keeps the relevant knowledge current as the work changes. A saved Module is available to later retrieval; it does not automatically replace context a CoAgent already loaded.

Keep boundaries explicit

Shared context is not a separate security boundary for every conversation. Give the CoAgent only the access appropriate for its role, and specify where results may be shared.

See Identity and access.