Skip to content

Other agents and editors

Any agent that speaks MCP can use the memory. For the common ones, pcx setup writes the server, the hooks and the instructions for you.

Sign in once, then set up the clients you use:

Terminal window
pcx login --token <your key>
pcx setup --client cursor,codex

By default the setup is global: written once into the client’s user-level config, it serves every project, and nothing is written into your repositories. Each project finds its own memory the same way the Claude Code plugin does.

Client What pcx setup writes
Codex CLI ~/.codex/config.toml, ~/.codex/AGENTS.md, ~/.codex/hooks.json, skills in ~/.agents/skills
Cursor ~/.cursor/mcp.json, ~/.cursor/hooks.json, ~/.cursor/skills
Gemini CLI ~/.gemini/settings.json, ~/.gemini/AGENTS.md, skills in ~/.agents/skills
Claude Desktop its user-level config
OpenCode, VS Code, Windsurf, Zed, Cline, Roo Code, Goose the project’s own files (these clients read them from the project)

To pin one project to fixed ids instead, and share them with your team:

Terminal window
pcx setup --project --client cursor,codex

The key is never written into a client file: the client runs pcx mcp, which forwards to the server with the key pcx login stored.

Some clients install from their own catalog. post-cortex ships as a Cursor plugin, a Gemini CLI extension, an entry in the MCP Registry, and native plugins for OpenCode, Kilo, Pi, OpenClaw and DeepSeek Harness. They take the server and key from their own settings, or from your pcx profile.

Point it at the MCP endpoint with your key as a bearer token:

URL https://api.pcx.is/mcp
Header Authorization: Bearer <your key>

For example, Claude Code without the plugin:

Terminal window
claude mcp add --transport http post-cortex https://api.pcx.is/mcp \
--header "Authorization: Bearer <your key>" --header "X-PCX-Author: claude-code"

X-PCX-Author names who writes; without it the server uses the client’s own name. pcx mcp-config <client> prints the entry for a client, ready to paste.