Local Worker MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| ENABLE_SHELL | No | Enable shell command execution | false |
| ALLOWED_PATHS | No | Comma-separated list of allowed paths for file access | |
| SECURITY_MODE | No | Security mode: READ_ONLY, WORKSPACE_WRITE, or FULL_LOCAL | READ_ONLY |
| CACHE_TTL_DAYS | No | Cache TTL in days | 30 |
| LOCAL_LLM_MODEL | No | The model name. If empty, auto-detects a model containing 'gemma' or uses the first available model | |
| LOG_MAX_SIZE_MB | No | Maximum log file size in MB before rotation | 250 |
| CACHE_MAX_SIZE_GB | No | Maximum cache size in GB | 10 |
| LOCAL_LLM_API_KEY | No | API key for OpenAI-compatible providers | |
| LOCAL_LLM_BASE_URL | No | The base URL of the LLM endpoint | http://127.0.0.1:11434 |
| LOCAL_LLM_PROVIDER | No | The LLM provider to use (ollama, openai_compatible, etc.) | ollama |
| LOG_RETENTION_DAYS | No | Number of days to retain log files | 14 |
| MAX_PARALLEL_WORKERS | No | Maximum number of parallel workers for batch delegation | 4 |
| LOCAL_LLM_MAX_RETRIES | No | Maximum number of retries | 0 |
| CACHE_TARGET_USAGE_PERCENT | No | Target cache usage percentage after cleanup | 80 |
| CACHE_CLEANUP_INTERVAL_HOURS | No | Interval in hours between automatic cache cleanups | 6 |
| CACHE_CLEANUP_THRESHOLD_PERCENT | No | Cache usage percentage threshold to trigger cleanup | 90 |
| LOCAL_LLM_CONNECT_TIMEOUT_SECONDS | No | Connection timeout in seconds | 2 |
| LOCAL_LLM_REQUEST_TIMEOUT_SECONDS | No | Request timeout in seconds | 45 |
| LOCAL_LLM_CIRCUIT_BREAKER_FAILURES | No | Number of failures before circuit breaker opens | 2 |
| LOCAL_LLM_CIRCUIT_BREAKER_COOLDOWN_SECONDS | No | Cooldown period in seconds before circuit breaker allows a retry | 60 |
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": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| local_statusA | Check local worker health: provider, endpoint, model, circuit breaker, cache. |
| delegate_taskC | Delegate a mechanical task to the local worker. Returns compact structured output. |
| delegate_batchB | Run independent local tasks in parallel. One failure does not cancel the batch. |
| delegate_fileC | Have the local worker read a file (txt/md/csv/json/code/log) without loading it into frontier context. |
| delegate_pdfC | Extract, chunk, and analyze a PDF locally. Returns compact findings with page evidence. |
| cache_statsB | Cache size, entry count, hits, misses, hit rate, and expired entries. |
| cache_cleanupA | Run cache garbage collection now (TTL, unused, then LRU). |
| cache_clearA | Delete disposable cache entries. Persistent artifacts stay unless include_persistent=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 8 tools
Each tool has a clear, distinct purpose: health check, task delegation, batch processing, file reading, PDF analysis, cache statistics, cleanup, and clearing. There is no overlap in functionality, making it easy for an agent to select the appropriate tool.
All tool names follow a consistent verb_noun pattern (e.g., local_status, delegate_task, cache_stats). The naming is predictable and clear, with verbs like 'delegate' and 'cache' used consistently to indicate the action and domain.
With 8 tools, the count is appropriate for the server's purpose of managing a local worker and its cache. The scope is well-defined without being too thin or overly broad.
The tool surface covers the key operations: health checks, task delegation (including batch and file processing), and cache management (stats, cleanup, clear). A minor gap is the lack of a tool to update or cancel a delegated task, but the core workflows are well supported.