ide-native mcp server protocol bridging
Exposes Theia IDE capabilities (file operations, editor state, workspace context) as a Model Context Protocol (MCP) server, enabling LLM clients to interact with the IDE through standardized MCP transport mechanisms (stdio, SSE, WebSocket). Implements MCP server specification with resource handlers, tool definitions, and prompt templates that map IDE operations to LLM-callable functions.
Unique: Bridges Theia IDE internals directly to MCP protocol without requiring custom LLM-specific plugins; leverages Theia's extension architecture to expose workspace/editor capabilities as standardized MCP resources and tools, enabling any MCP-compatible client to control the IDE
vs alternatives: More lightweight than building separate Theia plugins for each LLM provider; standardizes on MCP rather than proprietary IDE-LLM APIs, enabling tool reuse across Claude, Anthropic SDK, and other MCP hosts
workspace file and directory enumeration via mcp resources
Exposes the Theia workspace file tree as MCP resources, allowing LLM clients to list, read, and inspect directory structures and file metadata without direct filesystem access. Implements MCP resource handlers that traverse the workspace using Theia's FileService abstraction, supporting filtering by file type, size, and path patterns.
Unique: Leverages Theia's FileService abstraction to provide workspace enumeration via MCP, respecting IDE-level access controls and exclusion rules rather than raw filesystem access; integrates with Theia's virtual filesystem layer for remote/cloud workspaces
vs alternatives: More IDE-aware than raw filesystem APIs; respects workspace configuration and access controls; works seamlessly with remote Theia instances (cloud IDEs) where filesystem access isn't available
file read/write operations with editor state synchronization
Enables LLM clients to read and write files through MCP tools that integrate with Theia's editor state management. Writes trigger editor change events, update dirty state, and respect Theia's undo/redo stack. Reads return current editor content (including unsaved changes) rather than disk state, ensuring LLM sees what the user sees.
Unique: Integrates file operations with Theia's editor state machine, ensuring writes update the editor's dirty state and undo/redo stack; reads return editor buffer content (including unsaved changes) rather than disk state, providing LLM with accurate context
vs alternatives: More IDE-aware than raw file I/O; maintains consistency between LLM edits and editor state; respects Theia's change tracking and undo semantics unlike simple filesystem writes
editor cursor and selection state inspection
Exposes the current editor cursor position, text selection, and active editor context through MCP resources. Allows LLM clients to query which file is open, where the cursor is, and what text is selected, enabling context-aware code generation and refactoring targeted to specific locations.
Unique: Exposes Theia's editor selection model as queryable MCP resources, allowing LLM clients to understand user intent through cursor/selection context without requiring explicit user input
vs alternatives: Enables implicit context passing (LLM infers intent from selection) vs explicit prompting; tighter integration with IDE state than external LLM tools that don't have editor awareness
diagnostic and error reporting via mcp tools
Exposes Theia's diagnostic system (linter errors, type errors, warnings) as MCP resources and tools, allowing LLM clients to query problems in the workspace and receive structured error information. Integrates with Theia's MarkerService to surface language server diagnostics, build errors, and custom problem markers.
Unique: Bridges Theia's MarkerService and language server diagnostics to MCP, providing structured error context that LLM agents can use for intelligent code repair; integrates with Theia's diagnostic aggregation rather than re-running linters
vs alternatives: More efficient than LLM re-running linters; provides IDE-level error context that includes language server analysis; respects Theia's diagnostic filtering and severity levels
symbol definition and reference lookup via mcp
Exposes Theia's symbol navigation capabilities (go-to-definition, find-references, symbol outline) through MCP tools, allowing LLM clients to query code structure without parsing. Integrates with language servers to provide accurate symbol locations, type information, and cross-file references.
Unique: Delegates symbol resolution to Theia's language server integrations rather than implementing custom parsing; provides LLM with accurate, language-aware symbol information including type signatures and cross-file references
vs alternatives: More accurate than regex-based symbol search; language-aware (understands scoping, overloads, generics); leverages existing language server infrastructure rather than reimplementing symbol analysis
terminal command execution and output capture
Exposes Theia's integrated terminal as an MCP tool, allowing LLM clients to execute shell commands in the workspace context and capture output. Runs commands in the workspace directory with inherited environment variables, enabling agents to run build tools, tests, and custom scripts.
Unique: Integrates Theia's terminal service with MCP, enabling LLM agents to execute workspace commands and capture output; runs in workspace context with inherited environment, enabling tool chains (npm, python, etc.) to work seamlessly
vs alternatives: More integrated than external command execution; respects workspace environment and paths; enables AI agents to leverage existing build/test infrastructure without reimplementation
workspace configuration and settings inspection
Exposes Theia workspace settings, launch configurations, and extension configurations as MCP resources, allowing LLM clients to understand project setup and runtime environment. Provides access to .theia/settings.json, launch.json, and extension-specific configuration.
Unique: Exposes Theia's configuration system (including extension-specific settings) as queryable MCP resources, enabling LLM agents to understand project setup without parsing configuration files
vs alternatives: More complete than parsing config files manually; includes extension-specific settings and Theia-level configuration; respects Theia's configuration hierarchy (user/workspace/extension scopes)
+2 more capabilities