Connect ChatGPT
Install Socra Cortex from the ChatGPT plugin store, connect your Account, and add the instructions that teach ChatGPT when to use Cortex. You do not need a terminal or a custom Cloud Project for this path.
Open guided setup (new tab) to follow the preferred plugin path one step at a time.
Install the plugin
Open Socra Cortex in ChatGPT. Confirm the developer is socra and select Install plugin. If it is already installed, continue.
If your workspace blocks installation, ask its administrator to enable Cortex or choose another supported agent in guided setup (new tab).
Connect your Account
Before authorizing, review the access you are granting. Cortex tools can read and modify data that your selected Account permits. The read-only verification does not restrict the connection’s overall authority.
Installation or connection opens Socra sign-in in your browser. Select the same Account you use in Cortex, review access, and approve it there. If you already completed this during installation, continue without reconnecting.
If ChatGPT waits until first use to connect, start a chat, type @, select Socra Cortex, and ask it to list your Modules. Follow the sign-in prompt.
If no prompt appears, open Settings → Plugins → Socra Cortex and connect it. Signing in to the Cortex website alone does not authorize ChatGPT.
Enable all Cortex tools
We recommend Allow all for Cortex so ChatGPT can retrieve and maintain knowledge and coordinate work without repeated tool confirmations.
Open Settings → Plugins → Socra Cortex → Permission and select Allow all tools or Allow all actions, if available. Some accounts show Apps instead of Plugins. Keep the setting scoped to the Cortex app or connected account; availability depends on your account and workspace policy. If Connected accounts appears, open the intended Account’s ••• menu and choose Settings, then set its Permissions. Select Save if shown.
Allow-all covers the supported Cortex actions your Account permits, including writes and deletes. The suggested instructions guide the agent to maintain useful knowledge, name the affected Modules, and explain what changed and why. They do not enforce that behavior. If you prefer to approve tool calls, choose Always ask for Cortex.
For Claude, open Customize → Connectors → Socra Cortex → Tool permissions. Set both Read-only tools and Write/delete tools to Always allow using each group's menu. With the Cortex instructions in place, Claude can save useful knowledge, name the affected Modules, and explain what changed and why, without an extra tool-approval button. A custom connector may be named Cortex. In a new chat, make sure it is on under + → Connectors. Needs approval retains tool confirmations, and organization policy still applies.
ChatGPT's app-scoped controls were checked on September 26, 2026 against the signed-in interface and OpenAI's app-permission documentation. The Claude controls above were checked on September 5, 2026 against Claude connector permissions.
Add instructions
These suggested instructions teach your agent to find knowledge before acting and preserve useful decisions, preferences and reasons as you work. You can edit them, including asking for recommendations before Module changes are saved, or skip them. Connecting tools alone does not teach the agent when to retrieve and maintain knowledge.
Open Settings → Personalization → Custom instructions. You can also open Personalization from your profile menu. Add the following alongside your existing instructions and save if your host shows a Save action. On mobile, look for Customize ChatGPT and keep customization enabled.
Open just the Cortex instructions as Markdown. This link serves the current instructions without the installation guide.
Cortex helps people and teams stop repeating themselves to their AI. It keeps their preferences, requirements, decisions, and reasons in a single source of truth that agents can retrieve across conversations and tools.
Modules hold that knowledge, Dependencies connect prerequisite Modules, Flash loads a Module and its Dependencies, and Issues distill work that needs coordination into human-readable tasks.
Before executing a task, Flash its narrowest relevant Module(s). Flash known targets directly; otherwise prefer Search. Use Map only when Search leaves subjects or relationships unclear. Reuse a successful Flash while the required knowledge remains available in context. Retrieve missing knowledge again when needed.
When conversation or work reveals preferences, requirements, decisions, or other specific knowledge, update the relevant Modules or create missing ones. Preserve the knowledge and its reasons and keep Dependencies consistent. Briefly name the affected Modules and explain what changed and why.
These are the complete, unmodified instructions from the canonical Cortex session-start instructions. If the full text does not fit, use a project's instructions and work with Cortex in that project. Do not replace your other instructions. A message in one conversation does not configure future conversations.
For another MCP client, use its persistent or project instructions. For Claude, use Settings → Account → Instructions for Claude. Supported coding-agent plugins supply guidance through a skill or SessionStart hook, depending on the host. Follow the coding-agent guide for that host.
Check the connection
Start a new conversation with Socra Cortex enabled. Send:
Use Cortex to list the Modules I can access. Make a real Cortex tool call; an empty list is valid. Separately, report whether this new conversation received persistent Cortex guidance and what it requires. Do not change anything.
Check that ChatGPT calls Cortex and returns the Module list. An empty list is normal for a new Account. An access error or a response without a tool call does not confirm the connection; follow Troubleshoot Cortex.
Check the guidance report separately. If you saved instructions but the new conversation did not receive them, reopen Custom instructions to check the saved text and that customization is enabled. For project instructions, start the conversation inside that project. If you skipped instructions, the connection can still work; report the guidance as skipped.
Disconnect or remove Cortex
To stop using one connected Account, open Settings → Plugins → Socra Cortex → Connected accounts. Choose Actions for the exact named Account, then Disconnect. For a custom MCP plugin, open that exact plugin instead. Preserve your other connected accounts and plugins.
Removing the plugin is a separate action: open its supported plugin browser entry and choose Uninstall plugin when available. Workspace-managed plugins may require an administrator. OpenAI’s plugin documentation notes that separately connected MCP integrations remain connected until you disconnect them. A custom plugin may also offer Delete app; do not treat deletion or uninstall as proof that all authorizations are revoked.
For a manual MCP connection, disconnect the named ChatGPT account before separately removing any client you created for that connection through Custom MCP setup. Do not delete a shared client or another Account’s connection.
Keep working
Continue the work you already wanted to do. Share an idea, review a draft or correct a result. The agent can preserve useful specifics from that work without asking you to write a Module.
If you already have relevant material in the conversation, you can ask:
Use the decisions and corrections in this conversation to keep our Cortex knowledge current. Follow my review preference, check existing knowledge first, and tell me what changed and why.
This is optional; you do not need to prepare a knowledge library or complete an interview before starting. The agent can only use conversation history and artifacts available to it. The quickstart shows how later work can reuse what it learns.
The instructions guide the agent; they do not enforce its behavior. Choose the host confirmation settings that fit your needs.
For manual connections, use Custom MCP setup. For coding agents, follow Connect a coding agent.
Sources and verification
These controls were inspected in ChatGPT on October 13, 2026 for both the released Socra Cortex plugin and an existing custom Cortex MCP plugin. Disconnect was not executed in that inspection; its server-side revocation effects were not verified.