Connect a coding agent

Cortex plugins let your coding agent preserve and reuse knowledge as you work. You can start with an empty Account: the agent builds and maintains useful knowledge from decisions, corrections and material it is authorized to read. You do not need to write Modules yourself.

Each plugin provides Cortex tools and the guidance that teaches the agent when to retrieve, capture and update knowledge.

Cortex supports Codex, Claude Code, Cursor, and GitHub Copilot CLI.

Each plugin contains hosted MCP configuration. Codex receives guidance through the socra-cortex skill; its Cortex package has no SessionStart hook. Claude Code, Cursor, and GitHub Copilot CLI receive guidance through a SessionStart hook. For Grok Build, see Connect Grok.

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. Review later tool calls and approve only the actions you intend. The hosted MCP surface includes operations that can create, update, and delete Cortex resources. Review the security model before changing your agent host's approval policy.

Required for Claude Code, Cursor, and GitHub Copilot CLI: Approve or trust the Cortex SessionStart hook when the host asks. If the hook does not run, Cortex tools can remain available without the guidance it supplies.

Open guided setup (new tab) and choose Ask my agent to help. Your agent can follow the maintained setup guide; you complete sign-in and any consent your host requires. The steps below are also available if you prefer to install it yourself. For the store installation, follow Connect ChatGPT.

Install automatically

Use the Socra CLI to detect and configure every supported coding agent installed on your computer.

Install the Socra CLI

curl -fsSL https://cli.socra.cloud/install.sh | sh

irm https://cli.socra.cloud/install.ps1 | iex

Open a new terminal after the installer completes.

Install Cortex, sign in to your Account, and configure every detected coding agent:

socra install cortex
socra account login
socra cortex setup

The command detects supported coding agents, installs or updates each Cortex plugin, and starts supported authentication flows. The current Codex package needs no hook approval. Follow the command's next steps to finish any authorization that the other hosts perform inside their own interfaces.

Authentication required: Each coding agent stores its own Cortex authorization. socra cortex setup may start a browser login or report that authorization needs to be completed in the host. If Codex authentication is incomplete, run:

codex mcp login cortex

For Claude Code, start claude, run /mcp, select the Cortex server, and follow its browser sign-in. If setup reports an unsupported Claude login command, complete authentication this way.

For Cursor, open Customize → MCPs (older versions: Settings → Tools & MCP), select Cortex, and connect. For GitHub Copilot CLI, start copilot, run /mcp, select Cortex, and sign in. Complete Account selection in the browser that opens. If setup already connected the intended Account, continue without signing in again.

When a configured coding agent asks whether to allow the Cortex SessionStart hook, approve it.

You can run socra cortex setup again. It updates outdated installations and does not create duplicate installations.

Install manually

Use the instructions for your coding agent when you do not want the Socra CLI to manage plugin installation.

Codex

Add the Socra marketplace and install Cortex:

codex plugin marketplace add socra/agent-plugins --ref main
codex plugin add cortex@socra

Authenticate the MCP connection:

codex mcp login cortex

Select the Account you use in Cortex in the browser that opens. If Cortex is already connected to that Account, continue without logging in again.

Start a new Codex session. The Cortex plugin supplies the socra-cortex skill; this package has no SessionStart hook to approve.

Pause or remove Cortex in Codex

To pause the plugin’s Cortex server, set enabled = false in the existing [plugins."cortex@socra".mcp_servers.cortex] section of ~/.codex/config.toml. Preserve the section’s other settings and your other servers. Start a new session and use /mcp to check that Cortex is disabled. This pauses tool access; it does not establish server-side authorization revocation.

To remove the stored OAuth credentials for the Cortex HTTP server, run codex mcp logout cortex. To remove the plugin from your local Codex configuration and cache, run codex plugin remove cortex@socra. These are separate actions. Do not use codex logout, which signs Codex itself out. Follow OpenAI’s MCP policy documentation and CLI command reference; these commands act on local configuration and credentials, and do not establish server-side authorization revocation.

Claude Code

Add the Socra marketplace and install Cortex for your user profile:

claude plugin marketplace add socra/agent-plugins --scope user
claude plugin install cortex@socra --scope user

Authenticate the MCP connection:

Start claude, run /mcp, select the Cortex server, and follow the browser sign-in. Select the same Account that you use in Cortex. Already connected to the intended Account? Continue without authorizing again.

Approve the bundled Cortex SessionStart hook if Claude Code asks, then start a new Claude Code session.

Pause Cortex in Claude Code

To pause the Cortex server, toggle it off in /mcp. This preserves its configuration and applies to the current project. To disable the user-installed plugin and its SessionStart guidance hook, run claude plugin disable cortex@socra --scope user, then start a new session. Keep unrelated plugins enabled. These local controls do not establish server-side authorization revocation.

Cursor

Clone the public plugin repository and link the Cursor provider directory:

git clone https://github.com/socra/agent-plugins.git
mkdir -p ~/.cursor/plugins/local
ln -s "$(pwd)/agent-plugins/providers/cursor/cortex" ~/.cursor/plugins/local/cortex

Restart Cursor or run Developer: Reload Window. Confirm that Cortex appears under Customize. Approve the bundled Cortex SessionStart hook if Cursor asks. Open Customize → MCPs (older versions: Settings → Tools & MCP), select Cortex, and sign in if prompted. Choose the Account you use in Cortex. If it is already connected to that Account, continue without reconnecting. Keep your existing tool choices; make sure context_map is available for the first read-only check.

Pause Cortex in Cursor

To pause Cortex tools, turn off its server under Customize → MCPs. Preserve unrelated servers and your current tool choices. This local toggle does not establish server-side authorization revocation.

GitHub Copilot CLI

Add the Socra marketplace and install Cortex:

copilot plugin marketplace add socra/agent-plugins
copilot plugin install cortex@socra

Start copilot, approve the bundled Cortex SessionStart hook if GitHub Copilot CLI asks, then run /mcp. Select Cortex and sign in if prompted, choosing the same Account you use in Cortex. If it is already connected to that Account, continue without authorizing again. Start a new session after authorization completes.

Pause Cortex in GitHub Copilot CLI

Use copilot mcp list to identify the Cortex server and its plugin source. To pause that server, use copilot mcp disable with the listed server name, then start a new session. To disable the installed Cortex plugin and its SessionStart guidance hook, use copilot plugin disable with the plugin name shown by copilot plugin list. Preserve unrelated servers and plugins. These local controls do not establish server-side authorization revocation.

Other coding agents

Grok Build has a Cortex plugin. For another coding agent that supports remote HTTP MCP, configure the hosted Cortex MCP server. An MCP-only connection provides Cortex tools but does not install plugin guidance.

Allow the plugin instructions

The plugin supplies the suggested instructions. They teach your agent to Flash the right knowledge before acting and create or update Modules when you share useful knowledge or preferences. You do not need to add them to an AGENTS.md file. Verify guidance separately from tool access.

In Codex, start a new session with the Cortex plugin enabled and check that the socra-cortex skill is available. Claude Code, Cursor, and GitHub Copilot CLI may ask to trust the Cortex SessionStart hook during installation or startup; approve that hook and restart the session.

See the canonical Cortex instructions.

Enable Cortex tools

Check the Cortex server and the tools you intend to use:

• Codex: run /mcp and check that Cortex is connected. Keep context_map available for the first read-only Module-list check. You can retain a selected enabled_tools allowlist or disabled_tools list; wider tools are optional.
• Claude Code: run /mcp, select Cortex, and check that its server is connected. Keep context_map available for the first read-only Module-list check; wider tools are optional. Preserve your existing tool choices and confirmation rules.
• Cursor: open Customize → MCPs and confirm the connection. Keep context_map on for the first read-only Module-list check; wider tools in the list above chat are optional. Preserve your chosen tools and approval mode.
• GitHub Copilot CLI: run /mcp and check that Cortex is enabled and connected. Keep context_map available for the first read-only Module-list check; wider tools are optional. If the server was disabled, use copilot mcp list to identify it, then copilot mcp enable with that server name.

If you want fewer confirmation prompts, you can allow Cortex tools through your host's Cortex-scoped setting. Keeping normal confirmations is also supported:

• Codex: use default_tools_approval_mode = "prompt" under the existing [plugins."cortex@socra".mcp_servers.cortex] section in ~/.codex/config.toml to keep confirmations, or choose "approve" for fewer prompts. Keep enabled = true when you want to use Cortex and preserve your chosen tool lists. Approval policy and which tools are available are separate choices; per-tool overrides and organization policy still apply.
• Claude Code: keep confirmations, or optionally add mcp__plugin_cortex_cortex__* under /permissions in your user settings for future sessions. This Allow rule covers every tool from the plugin’s Cortex server; add it only if you want that approval policy. Remove the Cortex Allow rule to return to confirmations. Existing deny rules and organization policy still apply.
• Cursor: keep your chosen tools and approval mode, or optionally add cortex:* to the MCP allowlist under Settings → Agents → Approvals & Execution. This entry covers every Cortex tool, including write and delete operations. Preserve existing entries. If permissions.json controls the list, its mcpAllowlist overrides the in-app list; change only the Cortex entry if you want that policy. Cursor 3.6 and later can use Auto-review for non-allowlisted calls. Approval behavior depends on the selected mode and organization policy.
• GitHub Copilot CLI: start without an allow flag to keep normal confirmations, or optionally start with copilot --allow-tool='cortex'. The flag applies to that session and permits every tool from the Cortex server, including write and delete operations. Preserve your existing tool choices and approval policy; deny rules and organization policy still apply.

Broad Allow rules can skip tool confirmations for Cortex. The agent can save useful knowledge and tell you what changed and why. The plugin's session instructions supply this workflow without requiring an AGENTS.md file. Instructions guide behavior but do not enforce it, and existing deny rules or organization policy may still require confirmation or block a tool. You can retain normal confirmations instead.

You can also tell the agent to recommend Module changes and wait for your approval. That preference is separate from host tool confirmations. Either way, the agent prepares the knowledge and maintains its structure; you do not need to author it.

Current provider references: Codex MCP, Claude Code permissions, Cursor permissions, and Copilot CLI.

Verify the installation

Start a new session and ask:

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 both results: the coding agent makes a real Cortex call and returns the Module list, and the new session reports receiving Cortex guidance and explains what it requires. A new Account can return an empty list. If guidance is missing in Codex, check that the Cortex plugin and its socra-cortex skill are available. In Claude Code, Cursor, or GitHub Copilot CLI, check that the SessionStart hook is trusted and enabled. Start another session after correcting the configuration. Tool access alone does not confirm that the instructions loaded.

Next step

Continue your normal work with Cortex. The agent can start preserving useful knowledge as it appears; there is no graph to fill in first.

Sources and verification

The Claude Code controls above follow its official MCP reference, permissions guide and plugin commands. No host authorization, server toggle or plugin-disable action was executed in this verification; their effects on a live Cortex connection remain unverified.

Cursor controls follow its official MCP guide and permissions reference. Copilot controls follow GitHub’s CLI command reference and plugin reference. The older plugin --mcp flag has been removed; server controls now use the dedicated MCP commands. No host authorization, permission changes, server toggles or plugin-disable actions were executed for this verification.