RuvLTRA MCP Server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| LLAMA_CPP_PATH | No | Optional llama.cpp path hint | |
| RUVLTRA_CONFIG | No | Optional JSON config file path | |
| RUVLTRA_THREADS | No | llama.cpp thread count (0=auto) | 0 |
| RUVLTRA_LOG_LEVEL | No | debug/info/warn/error | info |
| RUVLTRA_GPU_LAYERS | No | llama.cpp GPU layers (-1 for auto) | -1 |
| RUVLTRA_HTTP_MODEL | No | HTTP model name | ruvltra-claude-code |
| RUVLTRA_MAX_TOKENS | No | Default max generation tokens | 512 |
| RUVLTRA_MODEL_PATH | No | Local GGUF model file path (auto-search if not set) | |
| RUVLTRA_HTTP_FORMAT | No | openai or llama | auto |
| RUVLTRA_MAX_WORKERS | No | Maximum worker count | 8 |
| RUVLTRA_MIN_WORKERS | No | Minimum worker count | 2 |
| RUVLTRA_TEMPERATURE | No | Default temperature | 0.2 |
| RUVLTRA_HTTP_API_KEY | No | HTTP API key | |
| RUVLTRA_RUVLLM_MODEL | No | RuvLLM auto-download model ID | ruvltra-claude-code |
| RUVLTRA_SONA_ENABLED | No | Enable SONA self-improvement | true |
| RUVLTRA_HTTP_ENDPOINT | No | HTTP inference endpoint (OpenAI-compatible or llama.cpp HTTP) | |
| RUVLTRA_CONTEXT_LENGTH | No | Context token length | 4096 |
| RUVLTRA_SONA_STATE_DIR | No | SONA state directory | ./.ruvltra-state/sona |
| RUVLTRA_HTTP_TIMEOUT_MS | No | HTTP timeout in milliseconds | 15000 |
| RUVLTRA_INITIAL_WORKERS | No | Initial worker count | 2 |
| RUVLTRA_MOCK_LATENCY_MS | No | Mock backend latency in milliseconds | 120 |
| RUVLTRA_TASK_TIMEOUT_MS | No | Default per-task timeout in milliseconds | 60000 |
| RUVLTRA_HTTP_MAX_RETRIES | No | HTTP retry count | 2 |
| RUVLTRA_QUEUE_MAX_LENGTH | No | Max queued tasks before backpressure | 256 |
| RUVLTRA_HTTP_RETRY_BASE_MS | No | Retry backoff base in milliseconds | 250 |
| RUVLTRA_SONA_PERSIST_INTERVAL | No | Persist every N interactions | 10 |
| RUVLTRA_HTTP_CIRCUIT_COOLDOWN_MS | No | Circuit cooldown in milliseconds | 30000 |
| RUVLTRA_HTTP_CIRCUIT_FAILURE_THRESHOLD | No | Failures before opening circuit | 5 |
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 | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| ruvltra_code_generateC | Generate code from instruction and context. For best output quality, provide instructions and context in English. |
| ruvltra_code_reviewB | Review code for bugs, security, performance, correctness, and maintainability. For best output quality, provide instructions and context in English. |
| ruvltra_code_refactorB | Refactor code while preserving behavior. For best output quality, provide instructions and context in English. |
| ruvltra_code_explainC | Explain code clearly for a given audience. For best output quality, provide instructions and context in English. |
| ruvltra_code_testC | Generate tests for provided code. For best output quality, provide instructions and context in English. |
| ruvltra_code_fixC | Fix code using error details and optional guidance. For best output quality, provide instructions and context in English. |
| ruvltra_code_completeB | Complete partial code from prefix/suffix. For best output quality, provide instructions and context in English. |
| ruvltra_code_translateB | Translate code between programming languages. For best output quality, provide instructions and context in English. |
| ruvltra_parallel_generateC | Generate multiple files concurrently through the worker pool. For best output quality, provide instructions and context in English. |
| ruvltra_swarm_reviewC | Run multi-perspective parallel code reviews (up to 8 perspectives). For best output quality, provide instructions and context in English. |
| ruvltra_statusB | Return server, worker, and backend runtime status. |
| ruvltra_sona_statsA | Return SONA learning statistics for all workers or a specific worker. |
| ruvltra_scale_workersA | Scale worker pool size within configured min/max. |
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 13 tools
Most tools have clearly distinct actions, but ruvltra_code_generate vs ruvltra_parallel_generate and ruvltra_code_review vs ruvltra_swarm_review could cause misselection if the single vs. parallel distinction is missed. Overall, the remaining boundaries are clear.
All tools share the ruvltra_ prefix and snake_case, and the code_* family follows a consistent pattern. Minor deviations exist with ruvltra_status and ruvltra_sona_stats being noun-like, and parallel_generate/swarm_review not following the code_* pattern.
13 tools is well within the ideal range and each tool has a distinct purpose in the code generation, review, and worker management lifecycle. The parallel and swarm variants are justified by the server's concurrency-oriented design.
The tool set covers the full code workflow: generate, review, refactor, explain, test, fix, complete, and translate, plus parallel variants. Operational coverage is also solid with status, stats, and worker scaling, leaving no obvious dead ends.