arcade-mcp/libs/arcade-mcp-server/docs/clients/inspector.md
Eric Gustin 3424ec8219
MCP Local (#563)
Versions:
* arcade-mcp\==1.0.0rc1
* arcade-mcp-server\==1.0.0rc1
* arcade-core\==2.5.0rc1
* arcade-tdk\==2.6.0rc1
* arcade-serve\==2.2.0rc1

### Summary
Adds first-class MCP support across Arcade, introduces a new MCP server
and CLI, unifies the project under the arcade-mcp name, overhauls
templates/scaffolding, and improves developer tooling, secrets
management, and examples.

### Highlights
- **MCP Server & Core**
- New MCP server with stdio and HTTP/SSE transports, session management,
resumability, and lifecycle handling.
- FastAPI-like `MCPApp` for building servers with lazy init; integrated
worker+MCP HTTP app option.
- Middleware system (logging and error handling), robust exception
hierarchy, and Pydantic-based settings.
- Async-safe managers for tools, resources, and prompts backed by
registries and locks.
- Developer-facing, transport-agnostic runtime context interfaces (logs,
tools, prompts, resources, sampling, UI, notifications).
- Conversion from Arcade ToolDefinition to MCP tool schema; OpenAI JSON
tool schema converter.
  - Parser supports `@app.tool`/`@app.tool(...)` decorators.

- **CLI**
  - New `mcp` command to run MCP servers with stdio or HTTP/SSE.
- New `secret` command to set/list/unset tool secrets (supports .env
input, preserves original casing for lookups).
- `new` command refactored; option to create a full toolkit package with
scaffolding.
  - `chat` command removed.
- `serve.py` imports updated to `arcade_serve.fastapi.telemetry`;
version retrieval now uses `arcade-mcp`.
  - `show.py` refactor to use new local catalog utilities.
- `display_tool_details` improved: adds “Default” column and handles
nested properties.

- **Configuration & Discovery**
- New `configure.py` to set up Claude Desktop, Cursor, and VS Code to
connect to local or Arcade Cloud MCP servers.
- Discovery utilities to find/install toolkits, build `ToolCatalog`s,
analyze files for tools, load kits from directories (pyproject parsing),
and build minimal toolkits.
- Better handling of provider API key resolution and evaluation suite
loading.

- **Templates & Scaffolding**
- Reorganized template structure (minimal vs full); moved
`.pre-commit-config.yaml`, `.ruff.toml`, license, Makefile, README,
tests, and tools layout to correct paths.
  - Minimal template adds `.env.example` for runtime secret injection.
- Template pyproject updated for MCP servers; includes sample server
with greeting and secret-reveal tools.
  - Authorization flow in templates simplified.

- **Repo-wide Renaming & Examples**
- Migrates references from `arcade-ai` to `arcade-mcp` across READMEs,
scripts, and package metadata.
- Examples updated (LangChain/LangGraph/AI SDK/TypeScript) and package
name changed to `arcade-mcp-sdk`.

- **Evals & Core Utilities**
- Evals now use OpenAI tooling format (`OpenAIToolList`, `to_openai`);
`tool_eval` takes `provider_api_key`.
- Core utilities: fixed `does_function_return_value` by dedenting before
parse; version bump to `2.5.0rc1` and dependency cleanup.

- **Tooling & CI**
- `setup-uv-env` action splits toolkit vs contrib dependency
installation.
- Pre-commit: excludes `libs/arcade-mcp-server/mkdocs.yml` and
`libs/tests/` from YAML and Ruff hooks; Ruff per-file ignores (e.g.,
C901 in `libs/**/*.py`, TRY400 in server docs paths).
- Makefile updates for uv env setup, quality checks, tests, builds, and
new `shell` target.
  - Added Makefile to MCP server library to streamline dev workflow.

- **Cleanup**
  - Removed `claude.json` config.
- Simplified stdio entrypoint; removed unused imports (`arcade_gmail`,
`arcade_search`).

### Breaking Changes
- **CLI**: `chat` command removed; use `mcp`, `secret`, and updated
`new`.
- **Naming**: All users should update references from `arcade-ai` to
`arcade-mcp`.
- **Templates**: File paths moved; downstream scripts referencing old
template locations may need updates.

### Getting Started
- Run an MCP server:
  - `arcade mcp --stdio --toolkits your_toolkit`
  - `arcade mcp --http --toolkits your_toolkit`
- Manage secrets:
  - `arcade secret set your_toolkit KEY=value`
  - `arcade secret list your_toolkit`
  - `arcade secret unset your_toolkit KEY`
- Configure clients:
- `arcade configure` to set up Claude Desktop, Cursor, and VS Code for
local/Arcade Cloud MCP.

---------

Co-authored-by: Sam Partee <sam@arcade-ai.com>
Co-authored-by: Shub <125150494+shubcodes@users.noreply.github.com>
2025-09-25 15:28:15 -07:00

354 lines
7 KiB
Markdown

# MCP Inspector
The MCP Inspector is a powerful debugging and testing tool for MCP servers. It provides a web-based interface to interact with your Arcade MCP server, test tools, and monitor protocol messages.
## Installation
Install the MCP Inspector globally:
```bash
npm install -g @modelcontextprotocol/inspector
```
Or use npx to run without installing:
```bash
npx @modelcontextprotocol/inspector
```
## Basic Usage
### Connecting to HTTP Servers
For MCP servers running over HTTP:
```bash
# Start your MCP server
python -m arcade_mcp_server --host 0.0.0.0 --port 8000
# In another terminal, start the inspector
mcp-inspector http://localhost:8000/mcp
```
### Connecting to stdio Servers
For stdio-based servers:
```bash
# Start the inspector with your server command
mcp-inspector "python -m arcade_mcp_server stdio"
# With additional arguments
mcp-inspector "python -m arcade_mcp_server stdio --tool-package github"
```
## Inspector Features
### Tool Explorer
The Tool Explorer shows all available tools with:
- Tool names and descriptions
- Parameter schemas
- Return type information
- Example invocations
### Interactive Testing
Test tools directly from the interface:
1. Select a tool from the explorer
2. Fill in parameter values
3. Click "Execute" to run the tool
4. View results and execution time
### Protocol Monitor
Monitor all MCP protocol messages:
- Request/response pairs
- Message timing
- Protocol errors
- Raw JSON data
### Resource Browser
If your server provides resources:
- Browse available resources
- View resource contents
- Test resource operations
### Prompt Templates
Test prompt templates if supported:
- View available prompts
- Fill template parameters
- Preview rendered prompts
## Advanced Usage
### Custom Environment
Pass environment variables to your server:
```bash
# Using env command
env ARCADE_API_KEY=your-key mcp-inspector "python -m arcade_mcp_server stdio"
# Using inspector's env option
mcp-inspector --env ARCADE_API_KEY=your-key "python -m arcade_mcp_server stdio"
```
### Working Directory
Set the working directory for your server:
```bash
mcp-inspector --cwd /path/to/project "python -m arcade_mcp_server stdio"
```
### Debug Mode
Enable verbose logging:
```bash
# Debug the MCP server
mcp-inspector "python -m arcade_mcp_server stdio --debug"
# Debug the inspector itself
mcp-inspector --debug "python -m arcade_mcp_server stdio"
```
## Testing Workflows
### Tool Development
1. **Start your server with hot reload**:
```bash
python -m arcade_mcp_server --reload --debug
```
2. **Connect the inspector**:
```bash
mcp-inspector http://localhost:8000/mcp
```
3. **Develop and test**:
- Modify your tool code
- Server auto-reloads
- Test immediately in inspector
### Performance Testing
Use the inspector to measure tool performance:
1. Enable timing in the Protocol Monitor
2. Execute tools multiple times
3. Analyze response times
4. Identify bottlenecks
### Error Debugging
Debug tool errors effectively:
1. Enable debug mode on your server
2. Execute the failing tool
3. Check Protocol Monitor for error details
4. View server logs in terminal
## Integration Testing
### Test Suites
Create test suites using the inspector:
```javascript
// test-tools.js
const tests = [
{
tool: "greet",
params: { name: "World" },
expected: "Hello, World!"
},
{
tool: "calculate",
params: { expression: "2 + 2" },
expected: 4
}
];
// Run tests via inspector API
```
### Automated Testing
Combine with testing frameworks:
```python
# test_mcp_tools.py
import subprocess
import json
import pytest
def test_tool_via_inspector():
# Start server
server = subprocess.Popen(
["python", "-m", "arcade_mcp_server"],
stdout=subprocess.PIPE
)
# Use inspector's API to test tools
# ...
```
## Best Practices
### Development Setup
1. **Use Split Terminal**:
- Terminal 1: MCP server with reload
- Terminal 2: Inspector
- Terminal 3: Code editor
2. **Enable All Debugging**:
```bash
python -m arcade_mcp_server --reload --debug --env-file .env.dev
```
3. **Save Test Cases**:
- Export successful tool calls
- Build regression test suite
- Document edge cases
### Production Testing
1. **Test Against Production Config**:
```bash
mcp-inspector "python -m arcade_mcp_server stdio --env-file .env.prod"
```
2. **Verify Security**:
- Test with limited permissions
- Verify API key handling
- Check error messages don't leak secrets
3. **Load Testing**:
- Execute tools rapidly
- Monitor memory usage
- Check for resource leaks
## Troubleshooting
### Connection Issues
#### "Failed to connect"
1. Verify server is running
2. Check correct URL/command
3. Ensure ports aren't blocked
4. Try with `--debug` flag
#### "Protocol error"
1. Ensure server implements MCP correctly
2. Check for version compatibility
3. Review server logs
4. Verify transport type
### Tool Issues
#### "Tool not found"
1. Verify tool is decorated with `@tool`
2. Check tool discovery in server
3. Ensure no import errors
4. Restart server and inspector
#### "Parameter validation failed"
1. Check parameter types match schema
2. Verify required parameters
3. Test with simpler values
4. Review tool documentation
## Examples
### Quick Test Session
```bash
# 1. Start a simple MCP server
cat > test_tools.py << 'EOF'
from arcade_mcp_server import tool
from typing import Annotated
@tool
def echo(message: Annotated[str, "Message to echo"]) -> str:
"""Echo the message back."""
return message
@tool
def add(
a: Annotated[int, "First number"],
b: Annotated[int, "Second number"]
) -> Annotated[int, "Sum"]:
"""Add two numbers."""
return a + b
EOF
# 2. Start inspector
mcp-inspector "python -m arcade_mcp_server stdio"
# 3. Test tools in the web interface
```
### HTTP Server Testing
```bash
# 1. Create an MCPApp server
cat > app.py << 'EOF'
from arcade_mcp_server import MCPApp
from typing import Annotated
app = MCPApp(name="test-server", version="1.0.0")
@app.tool
def get_time() -> Annotated[str, "Current time"]:
"""Get the current time."""
from datetime import datetime
return datetime.now().isoformat()
if __name__ == "__main__":
app.run(port=9000, reload=True)
EOF
# 2. Run the server
python app.py
# 3. Connect inspector
mcp-inspector http://localhost:9000/mcp
```
### Debugging Session
```bash
# 1. Enable all debugging
export DEBUG=*
export MCP_DEBUG=true
# 2. Start server with verbose logging
python -m arcade_mcp_server stdio --debug 2>server.log
# 3. Start inspector with debugging
mcp-inspector --debug "tail -f server.log" &
mcp-inspector --debug "python -m arcade_mcp_server stdio --debug"
```
## Tips and Tricks
1. **Bookmark Tool URLs**: Save frequently tested tools
2. **Export Test Data**: Save successful requests for documentation
3. **Use Browser DevTools**: Inspect network requests
4. **Create Tool Shortcuts**: Bookmark specific tool tests
5. **Monitor Resources**: Keep an eye on server resources during testing