UE5 MCP Server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MCP_SECRET | No | Adds X-MCP-Secret header to all Remote Control requests | |
| UE5_RC_URL | No | Remote Control API base URL | http://127.0.0.1:30010 |
| UE5_LOG_PATH | No | Absolute path to the .log file | |
| UE5_PROJECT_SOURCE | No | Absolute path to your project's Source/ directory | |
| UE5_STAGING_TTL_MS | No | Milliseconds before a staged batch expires | 600000 |
| UE5_ALLOW_FILE_WRITE | No | Must be exactly 'true' to enable file write tools | false |
| UE5_WRITE_EXTENSIONS | No | Comma-separated allowed extensions for file writes | .cpp,.h |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| ue5_run_console_commandA | Execute a console command in the running UE5 editor session. Use for stat commands, CVars, level operations, actor spawning, etc. Destructive commands (quit, open, exit) are blocked. |
| ue5_get_object_propertyA | Read a property value from any UObject in the current level by its full Unreal object path. |
| ue5_set_object_propertyB | Set a property value on any UObject in the current level. The value must match the expected JSON type for the property. |
| ue5_call_functionB | Call a Blueprint-callable UFUNCTION on any actor or component by its full Unreal object path. |
| ue5_search_sourceA | Search C++ and header files in the UE5 project Source directory using a regex pattern. Returns file path, line number, and surrounding context for each match. |
| ue5_read_fileA | Read the contents of a specific C++ source file in the project Source directory. Supports optional line range to limit output for large files. |
| ue5_read_logA | Read recent lines from the UE5 output log. Optionally filter to only Error or Warning severity. Reads from the tail of the file to avoid loading multi-GB logs. |
| ue5_stage_filesA | Stage one or more C++ files in memory for review before writing to disk. Nothing is written until ue5_commit_files is called. Requires UE5_ALLOW_FILE_WRITE=true. Only .cpp and .h files are permitted. |
| ue5_preview_stagedA | Preview the full content of all staged files before committing to disk. This is the review gate — inspect output before calling ue5_commit_files. |
| ue5_commit_filesA | Write all staged files (or a subset) to disk. Directories are created if needed. Optionally triggers a hot reload compile afterwards. Requires UE5_ALLOW_FILE_WRITE=true. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 10 tools
Each tool targets a distinct operation: console execution, object property read/write, function invocation, source search/read, log reading, and staged file editing. The only broad tool, run_console_command, is clearly framed as the engine command entry point and does not blur with the object or source tools.
All tools share the ue5_ prefix and follow lowercase snake_case with a leading verb: get, set, call, search, read, stage, preview, commit. The naming pattern is highly predictable and makes the tool surface easy to navigate.
Ten tools is well within the ideal range for a domain-specific MCP server. The set covers object inspection/mutation, console commands, source code access, log reading, and a safe file-writing workflow without unnecessary redundancy.
The core workflows are well covered, especially the staged file edit lifecycle and source/log access. The main gap is that object property and function tools require full Unreal object paths, but there is no dedicated tool for listing or discovering objects and actors; agents must rely on console commands as a workaround.