FastMCP Server

ignition-lint includes a FastMCP server that exposes linting capabilities to AI agents and MCP-compatible clients.

Prerequisites

Install the MCP extra:

pip install "ignition-lint-toolkit[mcp]"

Starting the Server

ignition-lint-server

The server starts and registers tools and resources that MCP clients can discover and invoke. The server itself is stateless — project paths are passed to individual tool calls rather than at startup.

Schema Modes

The server uses the robust schema mode by default. Three modes are available across the linter:

Mode	Behavior
strict	Rejects any property not in the schema
robust	Allows additional properties but validates known ones (default)
permissive	Minimal validation, accepts most structures

Available Tools

check_linter_status

Verify that the linting schema is available and report the current schema mode.

Returns: JSON with available, schema_mode, and schema_path fields.

lint_perspective_components

Lint Perspective view.json files in an Ignition project.

Parameter	Type	Required	Description
project_path	string	Yes	Path to the Ignition project root
component_type	string	No	Filter to a specific component type
verbose	boolean	No	Show detailed output
ignore_codes	string	No	Comma-separated rule codes to suppress

lint_jython_scripts

Lint Python/Jython scripts in the script-python directory.

Parameter	Type	Required	Description
project_path	string	Yes	Path to the Ignition project root
verbose	boolean	No	Show detailed output
ignore_codes	string	No	Comma-separated rule codes to suppress

lint_ignition_project

Run comprehensive linting across the entire project (Perspective, naming, scripts).

Parameter	Type	Required	Description
project_path	string	Yes	Path to the Ignition project root
lint_type	string	No	all, perspective, naming, or scripts (default: all)
component_type	string	No	Filter to a specific component type
verbose	boolean	No	Show detailed output
ignore_codes	string	No	Comma-separated rule codes to suppress

Available Resources

ignition://linter/status

Returns a JSON object with the current linter status, schema availability, and schema path.

ignition://linter/help

Returns a usage guide listing all available tools and their parameters.

Using --target with AI Agents

For AI agent and MCP workflows, the CLI's --target flag is the preferred interface. It accepts any directory and recursively discovers view.json and .py files without requiring a standard Ignition project layout:

# Lint any directory, JSON output for programmatic consumption
ignition-lint --target /path/to/any/folder --report-format json

# Lint only Perspective views in a subdirectory
ignition-lint -t /path/to/views/MyScreen --checks perspective --report-format json

# Lint only scripts
ignition-lint -t /path/to/scripts --checks scripts --report-format json

The JSON output includes structured issue data with file paths, severity, rule codes, component paths, and suggestions — ideal for agent-driven triage and auto-fix workflows.

Example: Calling from an MCP Client

# Using an MCP-compatible client
result = client.call_tool(
    "lint_ignition_project",
    project_path="/path/to/project",
    lint_type="all",
    ignore_codes="NAMING_PARAMETER,LONG_LINE",
)
print(result)

Suppression Support

All tool functions accept an optional ignore_codes parameter (comma-separated string). The .ignition-lintignore file in the project root is also read automatically. See the Suppression Guide for details.

New Rule Categories

The linter now includes additional checks surfaced through the MCP server:

• Expression rules (EXPR_*) — now() polling, unknown functions, malformed property refs, fragile component traversal
• Unused property rules (UNUSED_CUSTOM_PROPERTY, UNUSED_PARAM_PROPERTY) — dead custom/param properties per view
• Bad component ref (JYTHON_BAD_COMPONENT_REF) — fragile getSibling/getParent/getChild/getComponent usage in scripts

See the Rule Codes reference for the complete list.