# MCP Server Tool Evaluation Support
## Overview
Add support for evaluating tools from remote MCP servers without
requiring Python callables. Enables direct evaluation of any
MCP-compatible tool server.
## What's New
### Core Features
- **`MCPToolRegistry`**: Evaluate tools from a single MCP server
- **`CompositeMCPRegistry`**: Evaluate tools from multiple MCP servers
simultaneously
- **Automatic loaders**: `load_from_stdio()` and `load_from_http()` to
fetch tools from running servers
- **Automatic namespacing**: Tools prefixed with server name (e.g.,
`server_tool_name`)
- **Smart name resolution**: Use short names if unique, full names if
ambiguous
- **OpenAI strict mode**: Automatic schema conversion prevents parameter
hallucinations
### Usage
**Automatic Loading:**
```python
from arcade_evals import load_from_stdio, MCPToolRegistry
# Load tools automatically from MCP server
tools = load_from_stdio(["npx", "-y", "@modelcontextprotocol/server-github"])
registry = MCPToolRegistry(tools)
```
**Single MCP Server:**
```python
from arcade_evals import MCPToolRegistry, ExpectedToolCall
registry = MCPToolRegistry(mcp_tools)
suite = EvalSuite(catalog=registry)
suite.add_case(
expected_tool_calls=[
ExpectedToolCall(tool_name="tool_name", args={...})
]
)
```
**Multiple MCP Servers:**
```python
from arcade_evals import CompositeMCPRegistry, load_from_stdio
# Load from multiple servers
github_tools = load_from_stdio(["npx", "-y", "@modelcontextprotocol/server-github"])
slack_tools = load_from_stdio(["npx", "-y", "@modelcontextprotocol/server-slack"])
composite = CompositeMCPRegistry(
tool_lists={
"github": github_tools,
"slack": slack_tools,
}
)
suite = EvalSuite(catalog=composite)
suite.add_case(
expected_tool_calls=[
ExpectedToolCall(tool_name="github_list_issues", args={...})
]
)
```
## Implementation
### Files Changed
- **`libs/arcade-evals/arcade_evals/registry.py`** (NEW): Registry
abstractions and implementations
- **`libs/arcade-evals/arcade_evals/loaders.py`** (NEW): Automatic tool
loading from MCP servers
- **`libs/arcade-evals/arcade_evals/eval.py`** (MODIFIED): Enhanced
`ExpectedToolCall` and evaluation logic
- **`libs/arcade-evals/arcade_evals/__init__.py`** (MODIFIED): Exported
new registries and loaders
### Key Technical Details
- Added `BaseToolRegistry` interface for abstraction
- `MCPToolRegistry` handles single server tools
- `CompositeMCPRegistry` manages multiple servers with collision
detection
- `load_from_stdio()` and `load_from_http()` for automatic tool
discovery
- Fixed name normalization bug: MCP tools use underscores (not dots)
- Optimized tool copying: 2.5x faster via shallow copy
## Testing
- ✅ 41 tests passing (25 new tests added)
- ✅ `test_eval_mcp_registry.py`: MCPToolRegistry functionality
- ✅ `test_eval_composite_mcp.py`: CompositeMCPRegistry with multiple
servers
- ✅ Verified backward compatibility with Python tools
## Backward Compatibility
✅ **100% backward compatible** - No breaking changes
## Breaking Changes
**None**
<!-- CURSOR_SUMMARY -->
---
> [!NOTE]
> Adds end-to-end eval UX: examples, a robust CLI runner, and rich
outputs.
>
> - **New examples**: `eval_arcade_gateway.py`,
`eval_stdio_mcp_server.py`, `eval_http_mcp_server.py`,
`eval_comprehensive_comparison.py` with timeouts, error handling, and
track-based comparisons; detailed `README.md`
> - **CLI runner**: `arcade_cli/evals_runner.py` to execute
evals/capture in parallel with progress, error isolation, failed-only
filtering, context inclusion, and multi-provider/model support
> - **Output formatters**: `arcade_cli/formatters/` (txt, md, html,
json) for evals and capture; comparative and multi-model HTML with tabs
and context rendering
> - **Display refactor**: `display.py` now supports writing multiple
formats, failed-only disclaimers, include-context, and improved console
summaries
>
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
ff8acf9c34a6b61462a019a1ee9df081006517d0. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
---------
Co-authored-by: Francisco Liberal <francisco@arcade.dev>
Co-authored-by: Mateo Torres <torresmateo@gmail.com>
126 lines
4.2 KiB
Python
126 lines
4.2 KiB
Python
"""Tests for track management in comparative evaluations."""
|
|
|
|
import pytest
|
|
from arcade_evals._evalsuite._tool_registry import EvalSuiteToolRegistry
|
|
from arcade_evals._evalsuite._tracks import TrackManager
|
|
|
|
# Mark all tests in this module as requiring evals dependencies
|
|
pytestmark = pytest.mark.evals
|
|
|
|
|
|
class TestTrackManager:
|
|
"""Tests for TrackManager class."""
|
|
|
|
def test_create_track(self) -> None:
|
|
"""Test creating a new track."""
|
|
manager = TrackManager()
|
|
registry = EvalSuiteToolRegistry()
|
|
|
|
track_name = manager.create_track("Test Track", registry)
|
|
|
|
assert track_name == "Test Track"
|
|
assert manager.has_track("Test Track")
|
|
assert manager.track_count() == 1
|
|
|
|
def test_create_duplicate_track_raises(self) -> None:
|
|
"""Test that creating a duplicate track raises ValueError."""
|
|
manager = TrackManager()
|
|
registry1 = EvalSuiteToolRegistry()
|
|
registry2 = EvalSuiteToolRegistry()
|
|
|
|
manager.create_track("Track1", registry1)
|
|
|
|
with pytest.raises(ValueError, match="already exists"):
|
|
manager.create_track("Track1", registry2)
|
|
|
|
def test_get_registry(self) -> None:
|
|
"""Test getting a registry by track name."""
|
|
manager = TrackManager()
|
|
registry = EvalSuiteToolRegistry()
|
|
registry.add_tool({"name": "TestTool", "description": "Test"})
|
|
|
|
manager.create_track("MyTrack", registry)
|
|
retrieved = manager.get_registry("MyTrack")
|
|
|
|
assert retrieved is registry
|
|
assert retrieved.has_tool("TestTool")
|
|
|
|
def test_get_registry_nonexistent(self) -> None:
|
|
"""Test getting a nonexistent registry returns None."""
|
|
manager = TrackManager()
|
|
|
|
result = manager.get_registry("NonexistentTrack")
|
|
|
|
assert result is None
|
|
|
|
def test_get_track_names(self) -> None:
|
|
"""Test getting all track names."""
|
|
manager = TrackManager()
|
|
manager.create_track("Track1", EvalSuiteToolRegistry())
|
|
manager.create_track("Track2", EvalSuiteToolRegistry())
|
|
manager.create_track("Track3", EvalSuiteToolRegistry())
|
|
|
|
names = manager.get_track_names()
|
|
|
|
assert names == ["Track1", "Track2", "Track3"]
|
|
|
|
def test_get_track_names_empty(self) -> None:
|
|
"""Test getting track names when empty."""
|
|
manager = TrackManager()
|
|
|
|
names = manager.get_track_names()
|
|
|
|
assert names == []
|
|
|
|
def test_has_track(self) -> None:
|
|
"""Test checking if track exists."""
|
|
manager = TrackManager()
|
|
manager.create_track("Exists", EvalSuiteToolRegistry())
|
|
|
|
assert manager.has_track("Exists") is True
|
|
assert manager.has_track("DoesNotExist") is False
|
|
|
|
def test_track_count(self) -> None:
|
|
"""Test counting tracks."""
|
|
manager = TrackManager()
|
|
|
|
assert manager.track_count() == 0
|
|
|
|
manager.create_track("Track1", EvalSuiteToolRegistry())
|
|
assert manager.track_count() == 1
|
|
|
|
manager.create_track("Track2", EvalSuiteToolRegistry())
|
|
assert manager.track_count() == 2
|
|
|
|
def test_get_all_registries(self) -> None:
|
|
"""Test getting all registries."""
|
|
manager = TrackManager()
|
|
reg1 = EvalSuiteToolRegistry()
|
|
reg2 = EvalSuiteToolRegistry()
|
|
|
|
manager.create_track("Track1", reg1)
|
|
manager.create_track("Track2", reg2)
|
|
|
|
all_regs = manager.get_all_registries()
|
|
|
|
assert len(all_regs) == 2
|
|
assert all_regs["Track1"] is reg1
|
|
assert all_regs["Track2"] is reg2
|
|
|
|
def test_registries_are_isolated(self) -> None:
|
|
"""Test that each track has its own isolated registry."""
|
|
manager = TrackManager()
|
|
reg1 = EvalSuiteToolRegistry()
|
|
reg2 = EvalSuiteToolRegistry()
|
|
|
|
reg1.add_tool({"name": "Tool1", "description": "Tool for track 1"})
|
|
reg2.add_tool({"name": "Tool2", "description": "Tool for track 2"})
|
|
|
|
manager.create_track("Track1", reg1)
|
|
manager.create_track("Track2", reg2)
|
|
|
|
# Each registry only has its own tool
|
|
assert manager.get_registry("Track1").has_tool("Tool1")
|
|
assert not manager.get_registry("Track1").has_tool("Tool2")
|
|
assert manager.get_registry("Track2").has_tool("Tool2")
|
|
assert not manager.get_registry("Track2").has_tool("Tool1")
|