In every FITS file captured by the Seestar S50 telescope, there is a story of photons that have traveled for tens of millions of years, passing through dust and silence, arriving at a lens on Earth in the briefest moment of cosmic time. These ancient messages deserve more than manual processing—they deserve thoughtful automation.

But even the most beautiful light must pass through process.

🛠 The Architecture of Connection

The siril-mcp project is a Model Context Protocol (MCP) server that creates a bridge between AI models and Siril, the powerful open-source astronomical image processing software. Rather than replacing human expertise, it amplifies it—allowing models to understand, validate, and execute the well-established workflows that turn raw telescope data into cosmic art.

At its core, siril-mcp provides a structured interface for AI models to:

• Detect and validate Siril installations across platforms
• Check project structure and file organization
• Process Seestar S30/S50 telescope mosaics with appropriate filter settings
• Download and manage the latest processing scripts from the astrophotography community
• Execute calibration, registration, stacking, and stretching workflows

But like all good systems, siril-mcp begins with structure.

The server automatically validates project layouts, checking that light frames, calibration data, and output directories are organized according to Siril’s expectations. It leverages the excellent SSF scripts from @naztronomy, which guide Siril through the complete processing pipeline—from raw FITS files to polished astronomical images.

Where human effort once involved manual downloads, file organization, and script execution, the model can now understand project intent and execute the appropriate workflow with proper guardrails and validation.

💡 A Protocol for Understanding

The choice to build this as an MCP server isn’t about novelty—it’s about creating a proper interface between human intent and machine execution. Through the Model Context Protocol, AI models can now:

Understand Context: Check what files are present, validate project structure, and determine the appropriate processing workflow based on filter types (broadband vs narrowband).

Execute with Precision: Run Siril processing scripts with the correct parameters, handling both UV/IR block filters and light pollution filters appropriately.

Adapt to Environment: Automatically detect Siril installations across macOS, Linux, and Windows systems, whether installed as applications or from package managers.

Maintain Quality: Validate each step of the process, from binary detection to final mosaic output, with comprehensive error handling and logging.

This isn’t about replacing the astrophotographer’s eye or intuition. It’s about handling the repetitive, error-prone parts of the workflow so human attention can focus on the art and science of capturing the cosmos.

🌀 From Intent to Image

What draws me to this work is the elegance of the abstraction. An AI model can now say “process this Seestar project with broadband settings” and the MCP server handles:

• Validating the Siril installation and version
• Checking that the project directory contains the expected light frames
• Creating the appropriate SSF processing scripts automatically
• Executing the full calibration → registration → stacking → stretching pipeline
• Ensuring output directories are created and organized properly

The light in these images is real—it touched supernovae, passed through hydrogen clouds, scattered across interstellar ice. The processing pipeline should honor that journey with consistency, precision, and care.

🌌 Current Capabilities

siril-mcp is actively developed with these working features:

• ✅ Smart Detection - Automatic Siril binary detection across platforms with fallback strategies

• ✅ Version Management - Check installed Siril versions and compatibility

• ✅ Project Validation - Analyze project structure and identify missing components

• ✅ Mosaic Processing - Full Seestar S30/S50 telescope image processing pipeline

• ✅ Filter Intelligence - Supports both broadband (UV/IR) and narrowband (LP) workflows

• ✅ Script Automation - Auto-creates required SSF scripts from the naztronomy repository

• ✅ MCP Protocol - Full Model Context Protocol compliance with proper async handling

• 🔄 In Development - GUI integration for headless preprocessing, PyPI packaging, and extended telescope support

The project includes comprehensive testing—both Python unit tests and MCP integration tests—ensuring every tool works correctly and the protocol implementation is robust.

🌌 Next Steps

siril-mcp is still early. The roadmap includes:

• Structured prompts and memory for different processing goals (galaxies, nebulae, wide-field)

• More robust validation of FITS metadata and frame type separation

• Fine-grained control over siril CLI execution and output interpretation

• A feedback loop between the LLM and Siril logs for adaptive processing

In time, I hope it will serve not just as a tool for astrophotographers, but as a model for thoughtful automation—where machine learning augments curiosity, and infrastructure serves wonder.

---

← Back to projects