JCGE Agent Interface

JCGE MCP server for CGE discovery, guidance, examples, solving, validation, and reporting.

OtherJuliav0.3.1

JCGEAgentInterface

What is a CGE?

A Computable General Equilibrium (CGE) model is a quantitative economic model that represents an economy as interconnected markets for goods and services, factors of production, institutions, and the rest of the world. It is calibrated with data (typically a Social Accounting Matrix) and solved numerically as a system of nonlinear equations until equilibrium conditions (zero-profit, market-clearing, and income-balance) hold within tolerance.

What is JCGE?

JCGE is a block-based CGE modeling and execution framework in Julia. It defines a shared RunSpec structure and reusable blocks so models can be assembled, validated, solved, and compared consistently across packages.

What is this package?

MCP-compatible interface layer for agents that need to discover, load, solve, validate, and inspect

JCGE models. It exposes a small action API and a standard MCP stdio server over the same functionality.

The agent surface is intentionally focused on:

  • discovering installed JCGE packages and available functionality,
  • listing and describing reusable JCGEBlocks components,
  • guiding users through CGE model development with JCGE blocks,
  • guiding formulation choices such as equality systems, inequalities, MCP/complementarity, and optimization-style representations,
  • guiding source-data preparation with released JCGEImportData adapters while retaining model-owned SAM construction choices,
  • guiding calibration with the currently available JCGECalibrate loaders, SAM helpers, labeled containers, and parameter workflows,
  • guiding solver choice and reporting from generated equations,
  • listing released JCGEExamples reference model modules,
  • solving registered RunSpec models,
  • validating solved contexts,
  • rendering implemented equations, blocks, and symbols through JCGEOutput,
  • reporting or applying updates to released JCGE packages in the active Julia environment.

The package services are grouped in the documentation as discovery, modeling guidance, model interaction, reporting, environment maintenance, and MCP runtime services. See docs/src/services.md for the full service contract and current limits.

MCP server

The standard MCP transport is:

using JCGEAgentInterface
serve(transport=:mcp_stdio)

Minimal local MCP client configuration:

{
  "mcpServers": {
    "jcge": {
      "command": "julia",
      "args": [
        "--project=/path/to/JCGEAgentInterface.jl",
        "-e",
        "using JCGEAgentInterface; serve(transport=:mcp_stdio)"
      ]
    }
  }
}

The MCP tools use the jcge_ prefix:

  • jcge_capabilities
  • jcge_list_blocks
  • jcge_describe_block
  • jcge_modeling_guide
  • jcge_formulation_guide
  • jcge_solver_guide
  • jcge_import_data_guide
  • jcge_calibration_guide
  • jcge_reporting_guide
  • jcge_list_examples
  • jcge_package_status
  • jcge_update_packages
  • jcge_list_models
  • jcge_load_model
  • jcge_solve
  • jcge_validate_model
  • jcge_render_model
  • jcge_export_results

The original custom JSON action protocol remains available:

serve() reads JSON requests from stdin and writes JSON responses to stdout.

Docker and MCP Registry

OCI-oriented MCP registry metadata is declared in server.json.

Future version tags matching v*.*.* trigger .github/workflows/publish-mcp.yml, which:

  • verifies the tag version matches Project.toml,
  • builds the MCP Docker image,
  • pushes it to GitHub Container Registry as ghcr.io/equicirco/jcge-agentinterface-mcp:<release-version>,
  • updates server.json for that release,
  • publishes the metadata to the official MCP Registry through GitHub OIDC.

The Docker image includes the MCP ownership label required by the MCP Registry:

LABEL io.modelcontextprotocol.server.name="io.github.equicirco/JCGEAgentInterface.jl"

No release, tag, image publication, or registry publication is performed by editing this repository.

How to cite

If you use the JCGE framework, please cite:

Boero, R. JCGE - Julia Computable General Equilibrium Framework [software], 2026. DOI: 10.5281/zenodo.18282436 URL: https://JCGE.org

@software{boero_jcge_2026,
  title  = {JCGE - Julia Computable General Equilibrium Framework},
  author = {Boero, Riccardo},
  year   = {2026},
  doi    = {10.5281/zenodo.18282436},
  url    = {https://JCGE.org}
}

If you use this package, please cite:

Boero, R. JCGEAgentInterface.jl - MCP-compatible agent interface for JCGE tooling. [software], 2026. DOI: 10.5281/zenodo.18335121 URL: https://AgentInterface.JCGE.org SourceCode: https://github.com/equicirco/JCGEAgentInterface.jl

@software{boero_jcgeagentinterface_2026,
  title  = {JCGEAgentInterface.jl - MCP-compatible agent interface for JCGE tooling.},
  author = {Boero, Riccardo},
  year   = {2026},
  doi    = {10.5281/zenodo.18335121},
  url    = {https://AgentInterface.JCGE.org}
}

If you use a specific tagged release, please cite the version DOI assigned on Zenodo for that release (preferred for exact reproducibility).

Installation

Source-derived launch command. Check the maintainer’s required arguments and credentials before running:

bash
docker run -i --rm ghcr.io/equicirco/jcge-agentinterface-mcp:0.3.1

Set up in your AI client

Merge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.

json
{
  "mcpServers": {
    "io-github-equicirco-jcgeagentinterface-jl": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/equicirco/jcge-agentinterface-mcp:0.3.1"
      ]
    }
  }
}

Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.

Claude Desktop setup reference

Package

ghcr.io/equicirco/jcge-agentinterface-mcp:0.3.1docker

Compatible MCP Clients

JCGE Agent Interface works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.

  • Claude Desktop~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect.
  • Cursor~/.cursor/mcp.jsonRestart Cursor for changes to take effect.
  • VS Code.vscode/mcp.jsonReload VS Code window for changes to take effect.
  • Windsurf~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect.
  • Claude Code.mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.

Learn More