Skip to main content
Glama
gix9303
by gix9303
README.md
# EverOS Codex

[简体中文](./README.zh-CN.md)

Use a local [EverOS](https://github.com/EverMind-AI/EverOS) instance as long-term memory for Codex through MCP.

`everos-codex` provides:

- an MCP bridge for local EverOS
- repo-scoped default `project_id` inference
- `memory_search`, `memory_get`, `memory_save_turn`, `memory_flush`
- a bundled Codex skill: `everos-local-memory`

## Why

This project keeps EverOS in its native local shape:

- EverOS stays a local HTTP service, usually `http://127.0.0.1:8000`
- Codex connects through MCP over stdio
- memory stays local-first

## Install

Verify EverOS is running:

```bash
curl http://127.0.0.1:8000/health
```

Install the bundled Codex skill:

```bash
npx -y everos-codex install-skill
```

Add this MCP config to `~/.codex/config.toml`:

```toml
[mcp_servers.everos]
command = "npx"
args = ["-y", "everos-codex"]

[mcp_servers.everos.env]
EVEROS_BASE_URL = "http://127.0.0.1:8000"
EVEROS_APP_ID = "codex"
EVEROS_DEFAULT_PROJECT_ID = ""
EVEROS_AGENT_ID = "codex"
EVEROS_AGENT_NAME = "Codex"
EVEROS_USER_ID = "your-user-id"
EVEROS_USER_NAME = "Your Name"
EVEROS_TIMEOUT_MS = "60000"
```

Restart Codex or open a fresh thread.

## Quick checks

```bash
npx -y everos-codex doctor
npx -y everos-codex smoke
```

## CLI

```bash
everos-codex start
everos-codex doctor
everos-codex smoke
everos-codex smoke-e2e
everos-codex install-skill
```

## How project IDs work

If `project_id` is omitted and `EVEROS_DEFAULT_PROJECT_ID` is empty, `everos-codex` infers it from the current repo root name.

Examples:

- `water` -> `project_id=water`
- `store-ops-suite` -> `project_id=store-ops-suite`
- `EverOS-Codex` -> `project_id=everos-codex`

Explicit `project_id` still wins.

## Bundled skill

This repo bundles `skills/everos-local-memory`.

Its job is simple:

- search before repeating work
- save only high-value summaries
- verify new memory with `memory_get`
- avoid dumping whole conversations into memory

## Acknowledgement

This project is built on top of the open-source [EverMind-AI/EverOS](https://github.com/EverMind-AI/EverOS) project.

Thanks to the EverOS maintainers and contributors for making a local-first memory system available to the community.

## License

Apache-2.0

TDQS

A3.5/5.0

Scored across 4 tools

Disambiguation5/5

Each tool targets a distinct memory operation: flush clears session state, get lists memory entries, save_turn stores summaries, and search retrieves past context. No functional overlap.

Naming Consistency4/5

All tools follow the 'memory_<verb>' pattern, but 'save_turn' includes an object noun, making it slightly less consistent than the single-word verbs. Still, the pattern is clear and predictable.

Tool Count5/5

4 tools is ideal for a focused memory management server. Each tool serves a core function without redundancy or clutter.

Completeness4/5

The set covers memory flushing, retrieval, saving, and searching. Missing an explicit delete operation, but flush may serve that purpose. Minor gap, but core workflows are supported.

Maintenance

ActivityStale
ResponsivenessNo issues