Sbxclientapi #14
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,13 @@ | ||
| { | ||
| "name": "instavm-plugins", | ||
| "owner": { | ||
| "name": "InstaVM" | ||
| }, | ||
| "plugins": [ | ||
| { | ||
| "name": "instavm-coderunner", | ||
| "source": ".", | ||
| "description": "Execute Python code in a local sandboxed Apple container with fast kernel pool performance" | ||
| } | ||
| ] | ||
| } |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,10 @@ | ||
| { | ||
| "name": "instavm-coderunner", | ||
| "version": "1.0.0", | ||
| "description": "Execute Python code in a sandboxed Apple container via local CodeRunner MCP server", | ||
| "author": { | ||
| "name": "InstaVM" | ||
| }, | ||
| "homepage": "https://github.com/instavm/coderunner", | ||
| "license": "MIT" | ||
| } |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,8 @@ | ||
| { | ||
| "mcpServers": { | ||
| "instavm-coderunner": { | ||
| "command": "python3", | ||
| "args": ["${CLAUDE_PLUGIN_ROOT}/scripts/mcp-proxy.py"] | ||
| } | ||
| } | ||
| } |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,258 @@ | ||
| # InstaVM CodeRunner Plugin for Claude Code | ||
|
|
||
| This plugin enables Claude Code to execute Python code in a local sandboxed Apple container using [InstaVM CodeRunner](https://github.com/instavm/coderunner) via the Model Context Protocol (MCP). | ||
|
|
||
| ## Quick Start | ||
|
|
||
| ```bash | ||
| # 1. Install and start CodeRunner (one-time setup) | ||
| git clone https://github.com/instavm/coderunner.git | ||
| cd coderunner | ||
| sudo ./install.sh | ||
|
|
||
| # 2. Install the Claude Code plugin | ||
| claude plugin marketplace add github:BandarLabs/coderunner/instavm-coderunner-plugin | ||
| claude plugin install instavm-coderunner@instavm-plugins | ||
|
|
||
| # 3. Reconnect to MCP servers | ||
| /mcp | ||
| ``` | ||
|
|
||
| That's it! Claude Code now has access to all CodeRunner tools. | ||
|
|
||
| ## Installation | ||
|
|
||
| ### Prerequisites | ||
|
|
||
| **IMPORTANT:** You must have CodeRunner installed and running **before** installing this plugin. | ||
|
|
||
| #### Install CodeRunner | ||
|
|
||
| ```bash | ||
| git clone https://github.com/instavm/coderunner.git | ||
| cd coderunner | ||
| sudo ./install.sh | ||
| ``` | ||
|
|
||
| This will: | ||
| - Install Apple's container runtime (version 0.8.0+) | ||
| - Pull the `instavm/coderunner` container image | ||
| - Start the CodeRunner MCP service at `http://coderunner.local:8222/mcp` | ||
|
|
||
| **Verify CodeRunner is running:** | ||
| ```bash | ||
| curl http://coderunner.local:8222/execute -X POST -H "Content-Type: application/json" -d '{"code":"print(\"test\")"}' | ||
| ``` | ||
|
|
||
| ### Install via GitHub URL (Recommended) | ||
|
|
||
| ```bash | ||
| # Add the InstaVM marketplace from GitHub | ||
| claude plugin marketplace add github:BandarLabs/coderunner/instavm-coderunner-plugin | ||
|
|
||
| # Install the plugin | ||
| claude plugin install instavm-coderunner@instavm-plugins | ||
| ``` | ||
|
|
||
| ### Option 2: Install from Local Path | ||
|
|
||
| If you've cloned the repository locally: | ||
|
|
||
| ```bash | ||
| # Add local marketplace | ||
| claude plugin marketplace add /path/to/coderunner/instavm-coderunner-plugin | ||
|
|
||
| # Install the plugin | ||
| claude plugin install instavm-coderunner@instavm-plugins | ||
| ``` | ||
|
|
||
| ## Usage | ||
|
|
||
| Once installed, Claude will have access to the `execute_python_code` tool from CodeRunner. You can ask Claude to execute Python code: | ||
|
|
||
| ``` | ||
| Please execute this Python code: | ||
| ```python | ||
| import math | ||
| print(f"The square root of 16 is {math.sqrt(16)}") | ||
| ``` | ||
| ``` | ||
|
|
||
| Or simply: | ||
| ``` | ||
| Execute this code: print("Hello from CodeRunner!") | ||
| ``` | ||
|
|
||
| ## Available Tools | ||
|
|
||
| The plugin exposes the following MCP tools from CodeRunner: | ||
|
|
||
| - **execute_python_code**: Execute Python code in a persistent Jupyter kernel with full stdout/stderr capture | ||
| - **navigate_and_get_all_visible_text**: Web scraping using Playwright - navigate to URLs and extract visible text | ||
| - **list_skills**: List all available skills (both public and user-added) in CodeRunner | ||
| - **get_skill_info**: Get documentation for a specific skill (reads SKILL.md) | ||
| - **get_skill_file**: Read any file from a skill's directory (e.g., EXAMPLES.md, API.md) | ||
|
|
||
| ## Features | ||
|
|
||
| - **Local Execution**: Code runs on your machine in a sandboxed container | ||
| - **No Cloud Uploads**: Process files locally without uploading to cloud services | ||
| - **Fast Performance**: Pre-warmed Jupyter kernel pool (2-5 kernels) for quick execution | ||
| - **Full Output**: Returns stdout, stderr, execution time, and CPU time | ||
| - **Security**: Runs in Apple container isolation | ||
| - **MCP Integration**: Uses standard Model Context Protocol for tool communication | ||
|
|
||
| ## Configuration | ||
|
|
||
| By default, the plugin connects to `http://coderunner.local:8222/mcp` via a stdio proxy. To customize the URL, set the `MCP_URL` environment variable in `.mcp.json`: | ||
|
|
||
| ```json | ||
| { | ||
| "mcpServers": { | ||
| "instavm-coderunner": { | ||
| "command": "python3", | ||
| "args": ["${CLAUDE_PLUGIN_ROOT}/scripts/mcp-proxy.py"], | ||
| "env": { | ||
| "MCP_URL": "http://your-custom-url:8222/mcp" | ||
| } | ||
| } | ||
| } | ||
| } | ||
| ``` | ||
|
|
||
| ## Example Output | ||
|
|
||
| ```python | ||
| # Code: | ||
| print("Hello from CodeRunner!") | ||
| import time | ||
| time.sleep(0.1) | ||
| for i in range(3): | ||
| print(f"Count: {i}") | ||
| ``` | ||
|
|
||
| **Result:** | ||
| ``` | ||
| Hello from CodeRunner! | ||
| Count: 0 | ||
| Count: 1 | ||
| Count: 2 | ||
|
|
||
| Execution time: 0.156s | ||
| ``` | ||
|
|
||
| ## Requirements | ||
|
|
||
| - macOS 26.0+ (recommended) for Apple Container support | ||
| - Python 3.10+ | ||
| - CodeRunner installed and running | ||
| - Claude Code with MCP plugin support | ||
|
|
||
| ## Troubleshooting | ||
|
|
||
| ### Plugin shows "failed" status | ||
|
|
||
| **Most common cause:** CodeRunner container is not running. | ||
|
|
||
| **Solution:** | ||
| ```bash | ||
| # Check if CodeRunner is running | ||
| curl http://coderunner.local:8222/execute -X POST -H "Content-Type: application/json" -d '{"code":"print(\"test\")"}' | ||
|
|
||
| # If not running, restart it: | ||
| cd /path/to/coderunner | ||
| sudo ./install.sh | ||
| ``` | ||
|
|
||
| ### "Invalid Host header" error | ||
|
|
||
| The CodeRunner MCP server is running but rejecting requests. This means the container needs to be restarted with proper hostname configuration. | ||
|
|
||
| **Solution:** | ||
| ```bash | ||
| # Stop existing containers | ||
| sudo pkill -f container | ||
| container system start | ||
|
|
||
| # Restart CodeRunner container | ||
| container run --name coderunner --detach --rm --cpus 8 --memory 4g \ | ||
| --volume "$HOME/.coderunner/assets/skills/user:/app/uploads/skills/user" \ | ||
| --volume "$HOME/.coderunner/outputs:/app/uploads/outputs" \ | ||
| instavm/coderunner | ||
| ``` | ||
|
|
||
| ### MCP connection errors | ||
|
|
||
| 1. **Check MCP logs:** | ||
| ```bash | ||
| cat ~/Library/Caches/claude-cli-nodejs/-Users-manish-coderunner-instavm-coderunner-plugin/mcp-logs-*/latest/*.jsonl | ||
| ``` | ||
|
|
||
| 2. **Test MCP endpoint manually:** | ||
| ```bash | ||
| curl -H "Host: coderunner.local:8222" http://coderunner.local:8222/mcp \ | ||
| -X POST -H "Content-Type: application/json" \ | ||
| -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05"}}' | ||
| ``` | ||
|
|
||
| 3. **Verify proxy script:** | ||
| ```bash | ||
| cd /path/to/instavm-coderunner-plugin | ||
| python3 scripts/mcp-proxy.py | ||
| # Then send: {"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05"}} | ||
| ``` | ||
|
|
||
| ## How It Works | ||
|
|
||
| This plugin uses Claude Code's MCP server integration to connect to the local CodeRunner instance: | ||
|
|
||
| 1. The `.mcp.json` file defines the MCP server connection | ||
| 2. Python proxy script bridges stdio MCP to HTTP endpoint | ||
| 3. Tools from CodeRunner are automatically discovered and made available | ||
| 4. When you ask Claude to execute code, it uses the `execute_python_code` tool | ||
| 5. Results are returned via the MCP protocol | ||
|
|
||
| ## Publishing to GitHub Marketplace | ||
|
|
||
| This plugin is published on GitHub and can be installed directly via the GitHub URL. Here's how to publish updates: | ||
|
|
||
| ### 1. Commit and Push Changes | ||
|
|
||
| ```bash | ||
| cd instavm-coderunner-plugin | ||
| git add . | ||
| git commit -m "Update plugin" | ||
| git push origin main | ||
| ``` | ||
|
|
||
| ### 2. Users Install from GitHub | ||
|
|
||
| Users can install the plugin directly from GitHub: | ||
|
|
||
| ```bash | ||
| # Add the InstaVM marketplace | ||
| claude plugin marketplace add github:BandarLabs/coderunner/instavm-coderunner-plugin | ||
|
|
||
| # Install the plugin | ||
| claude plugin install instavm-coderunner@instavm-plugins | ||
| ``` | ||
|
|
||
| That's it! Claude Code will automatically pull the plugin files from GitHub. | ||
|
|
||
| ## Repository Structure | ||
|
|
||
| ``` | ||
| coderunner/ | ||
| └── instavm-coderunner-plugin/ # Plugin directory | ||
| ├── .claude-plugin/ | ||
| │ ├── marketplace.json # Marketplace metadata | ||
| │ └── plugin.json # Plugin manifest | ||
| ├── scripts/ | ||
| │ └── mcp-proxy.py # stdio-to-HTTP MCP proxy | ||
| ├── .mcp.json # MCP server configuration | ||
| └── README.md # This file | ||
| ``` | ||
|
|
||
| ## License | ||
|
|
||
| MIT |
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The
apt-get --fix-broken installcommand is removed. While it might seem like a cleanup, ensure that there are no broken packages that this command was intended to fix. If broken packages are expected, consider adding a comment explaining why this command is not needed or if the issue is handled elsewhere.