Pharo Smalltalk Interop Mcp Server
@mumez
MCP server to communicate local Pharo Smalltalk image.
概要
What is Pharo Smalltalk Interop Mcp Server?
A local MCP server that communicates with a Pharo Smalltalk image, providing code evaluation, introspection, search, package management, project installation, test execution, UI debugging, and server configuration. It is intended for developers working with Pharo Smalltalk.
How to use Pharo Smalltalk Interop Mcp Server?
Requires Python 3.10+, uv package manager, and a Pharo image with PharoSmalltalkInteropServer installed. Run using uvx (no installation) or after cloning the repository using uv run. Configure with environment variable PHARO_SIS_PORT (default 8086). The server exposes 22 MCP tools.
Key features of Pharo Smalltalk Interop Mcp Server
- Execute Smalltalk expressions and return results
- Retrieve class and method source code and comments
- Search classes, methods, traits, implementors, and references
- Export and import packages in Tonel format
- Install projects via Metacello
- Run test suites at package or class level
- Capture screenshots and inspect UI structure for debugging
- Retrieve and modify server settings dynamically
Use cases of Pharo Smalltalk Interop Mcp Server
- Debug and inspect a live Pharo image from an MCP-compatible client
- Automate package export/import and project installation in CI/CD pipelines
- Search for method implementors and references during refactoring
- Run tests and view results without leaving the development environment
- Assist AI‑powered coding assistants to interact with Pharo code
FAQ from Pharo Smalltalk Interop Mcp Server
What are the prerequisites?
Python 3.10 or later, uv package manager, and a Pharo image with PharoSmalltalkInteropServer installed.
How do I configure the server port?
Set the PHARO_SIS_PORT environment variable (default 8086).
Does the server work without a live Pharo instance?
No, a running Pharo image with PharoSmalltalkInteropServer is required. Tests use mocking to avoid this dependency.
What transports or authentication are supported?
The README only describes local execution via stdio (MCP default). No authentication or alternative transports are mentioned.
Can I add custom settings beyond the documented ones?
Yes, the apply_settings tool accepts arbitrary key‑value pairs; the documented setting is stackSize (default 100).
