verbalogicproject-creator

Codex Toolbox — Development skill for Claude Code

Development community

Build one useful MCP server while learning how Codex tools, skills, hooks, subagents and permissions fit together.

How to install Codex Toolbox

This entry records only its repository, not the path inside it, so there is no exact command to give. Open verbalogicproject-creator/codex-toolbox and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Codex Toolbox does

Build one useful MCP server while learning how Codex tools, skills, hooks, subagents and permissions fit together. Local SQLite FTS5 index of official docs; every search carries its source and retrieval date.

Alternatives in Development

  • Market Segments — Identify 3-5 customer segments with JTBD and product fit 7.8k ★
  • License — Check that every Python source file carries the required SPDX Apache-2.0 license header 1.1k ★
  • Claude Code Hooks — 🪝 A growing collection of useful Claude Code hooks 340 ★

README

Codex Toolbox Curriculum

Build one useful MCP server while learning how Codex tools, skills, hooks, subagents, plugins, permissions, and automation work together.

The running example is a local SQLite FTS5 index of official Codex documentation and the MCP specification. Every search carries its source and retrieval date. The core needs no API key. Live Codex exercises use your existing Codex account.

This is a Codex adaptation of [Claude-code-mcp](https://github.com/verbalogicproject-creator/Claude-code-mcp). See [provenance](NOTICE.md) and [actual verification evidence](docs/verification.md).

Start here

Requires Python 3.11+, SQLite with FTS5, and `uv`. Run from this folder:

uv sync --frozen
uv run python anchor-mcp/docs_mcp.py --data-dir .data/fixtures ingest --manifest exercises/fixtures/manifest.json --fixtures
uv run python anchor-mcp/docs_mcp.py --data-dir .data/fixtures query registration
uv run pytest -q

The fixture corpus is synthetic test data. For real questions, explicitly download the official corpus into a separate index:

uv run python anchor-mcp/docs_mcp.py --data-dir .data/live ingest
uv run python anchor-mcp/docs_mcp.py --data-dir .data/live query 'MCP' --corpus codex
uv run python scripts/configure.py --data-dir .data/live

The last command **prints** a configuration snippet. Review and merge it into your trusted project's `.codex/config.toml` to connect Codex. Start a fresh session and inspect `/mcp`. App and IDE clients on the same host share MCP configuration; different hosts need their own setup. Nothing here installs personal configuration automatically. See [setup](docs/03-anchor.md) and [plugin installation](docs/07-plugins.md).

Chapters

Chapter Working outcome
01 — Discover your environment Versioned CLI capability report
02 — Tools, guidance, permissions A bounded tool workflow and project instruction exercise
[03 — Hello MCP and