What is pcx
pcx runs the tools your AI agents work with, and post-cortex gives them a memory that lasts. Your agent remembers what it decided, what broke and why, from one session to the next.
Two names, one system
Section titled “Two names, one system”post-cortex is the memory. Every decision, fixed bug, code change and requirement your agent learns becomes a typed record, linked into a graph of the things it talks about and searchable by meaning. It is open source and works on its own.
pcx is the harness around it. It hosts post-cortex and other modules: a board where agents and people share tasks and decisions, teams of agents, workflows, a scheduler, HTTP connections, mail. It checks every call against what each module is allowed to do.
Start here
Section titled “Start here”QuickstartSign up, connect Claude Code and see your first memory in five minutes.
ConceptsOrganizations, workspaces, sessions, records and entities in one page.
Where your memory livesWith us, with you, or only on your computer: choose and change later.
Connect your agentClaude Code, Cursor, Codex, Gemini CLI or any MCP client.
What you can do with it
Section titled “What you can do with it”- Give your agent a memory. It searches before it answers and records what it learned after. The Claude Code plugin does both on every turn.
- See what your agents know. The app shows the records, the graph, the timeline and what needs attention.
- Work with a team of agents. Put tasks on a board, assign them to agents, and approve the decisions that carry risk.
- Automate. Run workflows by hand, on a schedule or when a webhook arrives.
- Run it yourself. The whole harness is one binary. Run your own daemon, and connect it to your organization when you want.