KX Plugins for Claude Code and Codex

Public marketplace of plugins for working with KX products: kdb+/q, KDB-X Python, KDB-X, KDB.AI, KX Insights Enterprise, KX Dashboards, and OneTick — for Claude Code and OpenAI Codex.

Each plugin packages one or more Skills — a folder of markdown that the agent loads automatically when it recognises a relevant task. Skills teach the agent repeatable workflows, quality standards, and domain expertise so everyone gets consistent, high-quality results. The layout is the same for every client — see Client support, below.

---

Plugins

Plugin	What it does	When it triggers
q-knowledge	kdb+/q language support — idiomatic q, qsql, IPC, kdb+ workflows; also /qlint-snippet for KX qlint	Writing q code, querying kdb+ tables, lint checks
pykx-knowledge	KDB-X Python — using kdb+/q from Python, type conversions, API guidance	Working with KDB-X and Python
kdbx-knowledge	KDB-X workflows, aimeta metadata authoring + discovery, and the KDB-X DB Service (setup, ingest, query, troubleshooting)	KDB-X platform, AI-native vector search, writing/reading aimeta annotations, DB Service ingest & queries
kdbai-knowledge	KDB.AI vector database — schema, hybrid search, AI integration; also sizing deployments (RAM/disk/GPU VRAM)	Building vector search or RAG with KDB.AI; sizing a KDB.AI deployment
kdbie-knowledge	KX Insights Enterprise — kxi CLI admin + Stream Processor pipelines	Using kxi to build/deploy packages, manage IE, or writing SP pipelines
kx-dashboards-knowledge	KX Dashboards — generate schema-valid dashboard JSON from natural language (core + per-component skills)	Building KX Dashboards, generating dashboard components/JSON
onetick-knowledge	Query OneTick market and reference data from Claude — MCP for SQL discovery/composition, onetick-py WebAPI wheel for execution over Arrow	Querying OneTick, composing/executing OneTick SQL, onetick-py/otp questions

---

Client support

Each plugin's source of truth is a canonical plugin.json (+ mcp.json if it has MCP servers) at the plugin root, in the Agent Plugins format — any client that supports the spec reads these directly (confirmed here for Codex; other spec adopters should work the same way). .claude-plugin/plugin.json and .mcp.json are generated from that source for Claude Code, which doesn't yet read the canonical files natively, and are overwritten on every regeneration — never hand-edit them. Skill content and directory layout don't change across clients. See tools/README.md for the full contract and how to regenerate.

.claude-plugin/marketplace.json is hand-maintained and read by both clients. Its plugin entries carry a policy block, which is a Codex field — Codex enforces it, and Claude Code ignores it at load time (claude plugin validate reports one expected Unknown field 'policy' warning per entry, and passes).

---

Other coding agents

This repo's plugins are authored to the Agent Plugins spec, so any client that supports it — Cursor, GitHub Copilot, and other adopters, in addition to Codex — should be able to install and read them directly, the same way Codex does (untested here beyond Codex; see Client support above). If you're on a coding agent that doesn't support the spec — Cline and others — Vercel's skills CLI (npx skills) can still pull individual SKILL.md files from any Git repo, including this one, into whichever agent-specific skills directory your tool expects:

npx skills add https://github.com/KxSystems/kx-skills.git

This installs skill content directly rather than through a plugin/marketplace install flow, so MCP server config and marketplace-level features (versioning, /plugin install/update) don't carry over — it's a good fit for pulling a single skill into an agent this repo doesn't otherwise support, not a replacement for the Claude Code / Codex installation paths above.

---

Installation

Pick the tab for your client.

Claude Code

Add the marketplace:

/plugin marketplace add KxSystems/kx-skills

Install plugins:

/plugin install q-knowledge@kx-skills
/plugin install pykx-knowledge@kx-skills
/plugin install kdbx-knowledge@kx-skills
/plugin install kdbai-knowledge@kx-skills
/plugin install kdbie-knowledge@kx-skills
/plugin install kx-dashboards-knowledge@kx-skills
/plugin install onetick-knowledge@kx-skills

Or browse interactively with /plugin and pick from the Discover tab.

Every plugin bundles an MCP server. Run /reload-plugins after installation to activate it without a full session restart.

Update later:

/plugin marketplace update kx-skills

Codex

Add the marketplace:

codex plugin marketplace add https://github.com/KxSystems/kx-skills.git

Install plugins:

codex plugin add q-knowledge@kx-skills
codex plugin add pykx-knowledge@kx-skills
codex plugin add kdbx-knowledge@kx-skills
codex plugin add kdbai-knowledge@kx-skills
codex plugin add kdbie-knowledge@kx-skills
codex plugin add kx-dashboards-knowledge@kx-skills
codex plugin add onetick-knowledge@kx-skills

Confirm an MCP-bundling plugin registered its server with codex mcp list.

Update later:

codex plugin marketplace upgrade kx-skills

---

Repo structure

Skill layout is unchanged. Each plugin also has a canonical plugin.json (source of truth) at its root; .claude-plugin/plugin.json is generated from it — see Client support, above.

kx-skills/
├── README.md                        ← you are here
├── LICENSE                          ← Apache-2.0
├── .claude-plugin/
│   └── marketplace.json             ← marketplace catalog
├── tools/                           ← canonical → Claude Code manifest generator
└── plugins/
    ├── q-knowledge/
    │   ├── plugin.json
    │   ├── README.md
    │   └── skills/
    │       ├── q/                   ← q language & kdb+ skill
    │       │   ├── SKILL.md
    │       │   ├── reference.md
    │       │   └── references/
    │       └── qlint-snippet/       ← KX qlint wrapper (executable skill)
    │           ├── SKILL.md
    │           └── scripts/run.sh
    ├── pykx-knowledge/
    │   ├── plugin.json
    │   └── skills/pykx/             ← KDB-X Python library
    │       ├── SKILL.md
    │       └── reference.md
    ├── kdbx-knowledge/
    │   ├── plugin.json
    │   ├── reference/               ← shared by kxmeta-* skills
    │   │   ├── agent-guide.md
    │   │   ├── meta.schema.json
    │   │   └── openapi.json
    │   └── skills/
    │       ├── kdbx/                ← KDB-X platform
    │       │   ├── SKILL.md
    │       │   ├── ai-reference.md
    │       │   └── gpu-reference.md
    │       ├── kxmeta-author/       ← writing aimeta annotations
    │       │   └── SKILL.md
    │       ├── kxmeta-discover/     ← probing aimeta at runtime
    │       │   └── SKILL.md
    │       └── kdbx-db-service/     ← KDB-X DB Service: setup, ingest, query
    │           ├── SKILL.md
    │           ├── recipes/
    │           └── references/
    ├── kdbai-knowledge/
    │   ├── plugin.json
    │   └── skills/
    │       ├── kdbai/                ← KDB.AI vector database
    │       │   ├── SKILL.md
    │       │   └── reference.md
    │       └── sizing/               ← deployment sizing (RAM/disk/GPU)
    │           ├── SKILL.md
    │           ├── reference.md
    │           └── scripts/estimate.py
    ├── kdbie-knowledge/
    │   ├── plugin.json
    │   └── skills/
    │       ├── kdbie-dev/          ← kxi CLI admin (packages, install, logs)
    │       │   └── SKILL.md
    │       └── pipeline-dev/       ← Stream Processor pipeline authoring
    │           └── SKILL.md
    ├── kx-dashboards-knowledge/    ← KX Dashboards NLX skill set (32 skills)
    │   ├── plugin.json
    │   ├── README.md
    │   └── skills/
    │       ├── README.md
    │       ├── kx-dashboard-core/  ← foundation skill — load first
    │       │   ├── SKILL.md
    │       │   └── generate.js     ← dashboard-JSON generator script
    │       ├── kx-datagrid/        ← one folder per component skill
    │       │   └── SKILL.md
    │       ├── kx-pivot-grid/
    │       │   ├── SKILL.md
    │       │   └── reference/      ← skill-specific reference assets
    │       └── ...                 ← 32 component skills in total
    └── onetick-knowledge/
        ├── plugin.json
        ├── mcp.json                 ← distinct MCP server (OneTick-Cloud, not kx-docs-mcp)
        ├── README.md
        └── skills/onetick-cloud/ ← OneTick SQL discovery + execution
            ├── SKILL.md
            └── scripts/
                └── onetick_exec.py

---

Contributing

We want this to grow. If you've found a better way to do something, improved a checklist, or want to add a new skill entirely — please open an issue or pull request against KxSystems/kx-skills.

Skills are plain markdown — no code required to improve them.

---

Principles

1. Skills are living documents. If a checklist item is wrong or missing, fix it.
2. Evidence over assertions. Skills must require proof of work, not just promises.
3. No gold-plating. Add complexity only when it prevents real problems.

---

License

Apache License 2.0 — see LICENSE.