Cortex MCP
Cortex keeps decisions and project context between Claude Code and Codex sessions in a local database and shows what it recalled
- 76 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Hooks record session activity and project context into local memory
- On first run the plugin installs its dependencies and starts PostgreSQL on 127.0.0.1
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/cortexDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks record session activity and project context into local memory
- On first run the plugin installs its dependencies and starts PostgreSQL on 127.0.0.1
This is third-party code. Review the repository files before installing.
What it does
Cortex is persistent memory for agents that runs locally: a SQLite file by default or PostgreSQL with pgvector if preferred. There is no LLM in retrieval, nothing leaves the machine unless you configure an integration, and no account or API key is needed. Project memory is a file you can delete. Claude Code and Codex share Cortex memory: both native plugins provide the same memory tools and 11 lifecycle hook modules for automatic capture, recall and project context. The same stdio server connects to Claude Desktop, Claude Cowork and other local MCP hosts, including Gemini CLI, Cursor, Windsurf and VS Code. The project states 57 memory tools, 36 implemented neuroscience-inspired mechanisms and a self-curating per-project wiki. The authors say plainly that committed ablations do not establish retrieval benefit of the mechanisms, and they publish no energy or CO2 figures. After install, dependencies are set up on the first session, about 2 minutes, about 5 with semantic search.
Who it is for. Developers who switch between Claude Code and Codex and want to keep decisions, fixes and project context locally.
Good fit when
- Yesterday architecture decision should surface in today session without a retelling
- Work alternates between Claude Code and Codex and memory has to be shared
- You need to inspect what retrieval returned before trusting an answer
- Project memory must not go to the cloud and has to sit in a file on disk
Not a fit when
- You need memory with no dependency install: the first session installs a set and it takes minutes
- The system is Ubuntu 22.04 or RHEL 9: the bundled PostgreSQL is not stated for them
- You need measured retrieval quality gains: the authors do not claim them
Example request
Recall which database schema decision we made last week and show where Cortex got itMCP
- Transport
- stdio
- Authentication
- not required
Limitations
- The first session after install sets up dependencies, about 2 minutes, about 5 with semantic search, and until ready edits go through without Cortex guards.
- Bundled PostgreSQL with pgvector on macOS and Linux with glibc 2.36 or newer, not on Ubuntu 22.04 or RHEL 9, with data in `~/.claude/methodology/managed-pg/`.
- The Claude Desktop `.mcpb` bundle carries the tools and has no hooks.
- The authors do not prove retrieval benefit for the 36 mechanisms, and the neuroscience references explain their design.
How to disable. Disable the plugin with `claude plugin disable hypermnesia-mcp@cortex-plugins`. Memory stays in `~/.claude/methodology/managed-pg/` and is deleted by hand.
FAQ
Is an account or API key needed?
No. Cortex runs locally and nothing leaves the machine unless you configure an integration.
Do Claude Code and Codex share one memory?
Yes, with the same configuration root and store; both plugins give the same memory tools.
How do I install it in Claude Code?
`claude plugin marketplace add cdeust/Cortex`, `claude plugin install hypermnesia-mcp`, then `claude --init-only` to install dependencies before the first session.
Does Cortex work outside Claude Code and Codex?
Yes, the stdio server connects to Claude Desktop, Claude Cowork, Gemini CLI, Cursor, Windsurf and VS Code.
Where is the data stored?
By default in a local SQLite file; the bundled PostgreSQL keeps data in `~/.claude/methodology/managed-pg/`.
Discussion
No questions or reviews yet.
Related
MCP serversMemory for agents: a Claude Code session resumes in Codex with the same decisions and open tasks.
Shared memory for coding agents: messages in SQLite, old context compacted into a summary tree and MCP tools to search it.
Agent memory and a knowledge graph in one SQLite file: the plugin adds /remember and /recall to Claude Code and Codex.
hipocampus gives an agent file-based memory: a 3,000-token topic index loads each session and shows what came up before
Mnemo Cortex keeps agent memory locally: it recalls by meaning, compacts sessions, shares between agents and connects through MCP.
Your own server gives every agent shared memory, AST-level code search and pre-commit checks through one MCP endpoint.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Cortex (this card) | no | not required | 76 | |
| Claude-Mem | no | 116,457 | a persistent memory plugin for Claude Code, while Cortex shares memory between Claude Code and Codex and shows what was recalled | |
| CortexDB | no | 480 | memory and a knowledge graph in one SQLite file with `/remember` and `/recall`, while here are 57 tools and lifecycle hooks |
README in short
The README opens with the claim that everything runs locally and a section on shared Claude Code and Codex memory, then install for Claude Code, Claude Desktop, Cowork and Codex, launch commands for other MCP hosts and dependency details. The verification section spells out what measurements do not establish.