https://github.com/run-llama/llamaparse-agent-plugins
Claude Code/Codex plugins for LlamaParse, LiteParse and LlamaParse MCP
https://github.com/run-llama/llamaparse-agent-plugins
agent-skills claude-code claude-code-plugin codex codex-plugin mcp
Last synced: 1 day ago
JSON representation
Claude Code/Codex plugins for LlamaParse, LiteParse and LlamaParse MCP
- Host: GitHub
- URL: https://github.com/run-llama/llamaparse-agent-plugins
- Owner: run-llama
- License: mit
- Created: 2026-05-06T13:16:34.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2026-07-02T10:41:42.000Z (9 days ago)
- Last Synced: 2026-07-10T15:33:15.660Z (1 day ago)
- Topics: agent-skills, claude-code, claude-code-plugin, codex, codex-plugin, mcp
- Language: Python
- Homepage:
- Size: 135 KB
- Stars: 2
- Watchers: 0
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# LlamaParse Agent Plugins Marketplace
This repository contains local Codex and Claude plugin marketplace definitions with three document-focused plugins from LlamaIndex.
## Repository Contents
- `.agents/plugins/marketplace.json`: Marketplace definition consumed by Codex.
- `.claude-plugin/marketplace.json`: Marketplace definition consumed by Claude.
- `plugins/liteparse`: Local OCR parsing plugin (no cloud dependency).
- `plugins/llamaparse`: Cloud parsing skill plugin.
- `plugins/llamaparse-mcp`: LlamaParse Platform MCP plugin (tools for parse/split/classify plus Index v2 retrieval via MCP).
## Plugins Included
1. `liteparse`
- Type: Skill plugin
- Purpose: Parse and extract content from PDFs, Office docs, and images locally.
2. `llamaparse`
- Type: Skill plugin
- Purpose: Cloud-based, advanced document parsing.
3. `llamaparse-mcp`
- Type: MCP + skills plugin
- Purpose: Connects Codex to LlamaParse Platform MCP tools (document processing and Index v2 retrieval).
- Skills: `llamaparse-mcp` (parse/split/classify usage guide) and `llamacloud-index` (agentic retrieval over Index v2 knowledge bases).
- MCP server config: `plugins/llamaparse-mcp/.mcp.json` -> `https://mcp.llamaindex.ai/mcp`
- (Codex) Auth policy in marketplace: `ON_USE`
## Install The Marketplace
### Codex
```bash
codex plugin marketplace add run-llama/llamaparse-codex-plugins
```
### Claude
(within Claude)
```bash
/plugin marketplace add run-llama/llamaparse-agent-plugins
```
## Enable Plugins
### Codex
1. Start or reload Codex
2. Open plugin management in Codex.
3. Find marketplace `llamaparse-marketplace` (display name: `LlamaParse Codex MarketPlace`).
4. Install any of the following plugins:
- `liteparse`
- `llamaparse`
- `llamaparse-mcp`
**Authentication Notes**
- `liteparse`: No authentication required, runs fully locally.
- `llamaparse-mcp`: Prompts on first use (`ON_USE`) when invoking MCP tools.
- `llamaparse`: Follow any in-app prompts for required credentials ([LlamaParse API key](https://cloud.llamaindex.ai/signup)).
**Verify Plugin Availability**
After enabling plugins, confirm they are available in a Codex thread:
- `liteparse` skill should appear as `liteparse:liteparse`.
- `llamaparse-mcp` skills should appear as `llamaparse-mcp:llamaparse-mcp` and `llamaparse-mcp:llamacloud-index`.
- MCP tools should include operations like:
- `parseFile`
- `splitFile`
- `classifyFile`
- `uploadFileByUrl`
- `listIndexes`
- `retrieveFromIndex`
### Claude
Within Claude, run:
```bash
/plugin install liteparse@llamaparse-marketplace
/plugin install llamaparse@llamaparse-marketplace
/plugin install llamaparse-mcp@llamaparse-marketplace
```
**Use plugins**
Invoke the plugins as slash commands directly within Claude:
```bash
/liteparse:liteparse # -> run LiteParse skill
/llamaparse:llamaparse # -> run LlamaParse skill
/llamaparse-mcp:llamaparse-mcp # -> enable LlamaParse MCP and its document-processing skill
/llamaparse-mcp:llamacloud-index # -> enable LlamaParse MCP and its Index v2 agentic-retrieval skill
```
## Quick Smoke Test (LlamaParse MCP)
1. Request an upload URL with `getUploadUrl`.
2. Upload a file (for example `test.txt`) to the returned pre-signed endpoint.
3. Call `parseFile` with the returned `fileId`.
Expected result: parsed text content is returned.
## Quick Smoke Test (Index v2)
1. Call `listIndexes` to confirm at least one index is visible.
2. Call `retrieveFromIndex` with a natural-language question against one of the returned indexes.
Expected result: relevant passages from your indexed documents are returned.
## Development Notes
- Plugin manifests are located at:
+ `plugins/liteparse/.*-plugin/plugin.json`
+ `plugins/llamaparse/.*-plugin/plugin.json`
+ `plugins/llamaparse-mcp/.*-plugin/plugin.json`
- MCP server mapping is in `plugins/llamaparse-mcp/.mcp.json`.
### Bump version
Bump the version of a plugin for a specific agent with:
```bash
python3 scripts/bump-version.py --bump
# if you have uv
./scripts/bump-version.py --bump
```
For example:
```bash
./scripts/bump-version.py codex liteparse --bump minor
./scripts/bump-version.py claude liteparse --bump minor
```