Document-intelligence plugins for coding agents. Install any of the five into Claude Code, Codex CLI, Cursor, Gemini CLI, Factory Droid, GitHub Copilot CLI, or opencode.
| Plugin | Value Proposition | Status |
|---|---|---|
| xberg | Local document extraction from 91+ formats (PDF, Office, images with OCR, HTML, email, archives, academic) | Stable — v0.2.2 |
| crawlberg | Web crawling and scraping with HTML→Markdown and headless-Chrome fallback | Stable — v0.2.2 |
| html-to-markdown | Fast, lossless HTML→Markdown with structured metadata and tables | Stable — v0.2.2 |
| liter-llm | Universal LLM API client for 143 providers (chat, streaming, tools, embeddings) | Stable — v0.2.2 |
| tree-sitter-language-pack | Parse and extract code intelligence from 300+ languages | Stable — v0.2.2 |
Claude Code
Once approved by the marketplace:
/plugin install xberg@claude-community
/plugin install crawlberg@claude-community
/plugin install html-to-markdown@claude-community
/plugin install liter-llm@claude-community
/plugin install tree-sitter-language-pack@claude-community
Self-host (works today):
/plugin marketplace add xberg-io/plugins
/plugin install xberg@xberg
/plugin install crawlberg@xberg
/plugin install html-to-markdown@xberg
/plugin install liter-llm@xberg
/plugin install tree-sitter-language-pack@xberg
Pending review for official Claude marketplace.
Codex CLI
Self-host install:
/plugins add https://github.com/xberg-io/plugins
Then search for the plugin you want — e.g. xberg, crawlberg, html-to-markdown, liter-llm, or tree-sitter-language-pack — and select "Install Plugin".
Cursor
Self-host install only:
Settings → Plugins → Add from URL → https://github.com/xberg-io/plugins. Select the plugin(s) you want.
Gemini CLI
Self-host install:
gemini extensions install https://github.com/xberg-io/plugins
Factory Droid
Self-host install:
droid plugin marketplace add https://github.com/xberg-io/plugins
droid plugin install xberg@xberg
droid plugin install crawlberg@xberg
droid plugin install html-to-markdown@xberg
droid plugin install liter-llm@xberg
droid plugin install tree-sitter-language-pack@xberg
Pending review for official Factory Droid marketplace.
GitHub Copilot CLI
Self-host install:
copilot plugin marketplace add https://github.com/xberg-io/plugins
copilot plugin install xberg@xberg
copilot plugin install crawlberg@xberg
copilot plugin install html-to-markdown@xberg
copilot plugin install liter-llm@xberg
copilot plugin install tree-sitter-language-pack@xberg
opencode
Add the published packages to opencode.json:
{
"$schema": "https://opencode.ai/config.json",
"plugin": [
"@xberg-io/opencode-xberg",
"@xberg-io/opencode-crawlberg",
"@xberg-io/opencode-html-to-markdown",
"@xberg-io/opencode-tree-sitter-language-pack"
]
}liter-llm has no opencode package.
Each plugin shells out to a real CLI. Install whichever you use:
| Plugin | Binary | Install |
|---|---|---|
| xberg | xberg |
brew install xberg-io/tap/xberg |
| crawlberg | crawlberg |
brew install xberg-io/tap/crawlberg |
| html-to-markdown | html-to-markdown |
brew install xberg-io/tap/html-to-markdown |
| liter-llm | liter-llm |
brew install xberg-io/tap/liter-llm |
| tree-sitter-language-pack | tree-sitter-language-pack |
brew install xberg-io/tap/tree-sitter-language-pack |
Each plugin ships SKILL.md files describing what it can do. Agent harnesses auto-load skills based on the description: frontmatter in each file. When you ask your agent to extract a document or crawl a site, the matching skill fires automatically — you don't invoke skills directly.
Example: when you say "extract text and tables from this PDF", the xberg skill detects the request and loads the extract MCP tool from the local xberg binary. The agent then calls that tool with your document, getting back structured text, tables, and metadata. The same pattern applies to web crawling with crawlberg.
Skills are loaded at agent startup. Their descriptions stay in context so agents can decide when to use them. This means skills consume zero tokens unless the agent decides to invoke them.
.claude-plugin/
├── marketplace.json # Claude marketplace
.github/plugin/
├── marketplace.json # Copilot CLI marketplace
plugins/
├── xberg/
│ └── plugin.json # MCP server config
└── crawlberg/
└── plugin.json
scripts/
├── bump-version.sh # version bump/sync
└── validate-manifests.sh # CI parity check
Two independent version lines:
- Marketplace plugins (Claude / Codex / Cursor / Factory / GitHub / Gemini manifests) track
VERSION. - opencode npm packages (
@xberg-io/opencode-*) trackOPENCODE_VERSIONand publish to npm.
bump-version.sh writes the version file and every manifest in that group when given a version. Without a version, it syncs the group's manifests from its source-of-truth version file. validate-manifests.sh checks both groups.
scripts/bump-version.sh 0.2.3 # marketplace plugins (default group)
scripts/bump-version.sh --group opencode 0.1.1 # opencode npm packages
scripts/bump-version.sh --group opencode # resync from OPENCODE_VERSION
scripts/validate-manifests.sh
git commit -am "chore: release"
git tag vX.Y.Z && git push --tagsThe opencode packages publish from .github/workflows/publish.yaml (npm Trusted Publishing, with provenance) when a GitHub release is published. The first release was done locally with scripts/publish-opencode.sh.
Contributions are welcome. See CONTRIBUTING.md for guidelines on adding skills, testing locally, and bumping versions.
Report security vulnerabilities privately at SECURITY.md. Do not open public issues for security concerns.
Can I use just one plugin?
Yes. /plugin install xberg@xberg installs only xberg.
Do skills consume context?
Only when triggered. Skill descriptions load at startup (~300 tokens total) but stay in context. They consume zero tokens unless the agent decides to invoke the MCP tools.
Can I customize a skill locally?
Yes. Fork the repo, edit the skill SKILL.md, and use /plugin marketplace add <local-path> to self-host your version.
How do I report a broken skill?
Open an issue at xberg-io/plugins with the plugin name, agent harness, and exact request that failed.
MIT. See LICENSE.
- Xberg — document intelligence: text, tables, metadata from 91+ formats with optional OCR.
- Xberg Enterprise — managed extraction API with SDKs, dashboards, and observability.
- crawlberg — web crawling and scraping with HTML→Markdown and headless-Chrome fallback.
- html-to-markdown — fast, lossless HTML→Markdown engine.
- liter-llm — universal LLM API client with native bindings for 14 languages and 143 providers.
- tree-sitter-language-pack — tree-sitter grammars and code-intelligence primitives.
- alef — the polyglot binding generator that produces every per-language binding across the 5 polyglot repos.