Terradev MCP server

• active_contextFree

  Get current Terradev state: running training jobs, active instances, spend-to-date, alerts. Call this on session start to resume context from previous sessions.

• agent_agentic_serving_configureFree

  Configure agentic inference serving settings.

• agent_agentic_serving_helm_valuesFree

  Print Helm values for agentic inference deployment.

• agent_agentic_serving_k8sFree

  Print K8s deployment manifests for agentic inference.

• agent_agentic_serving_launch_argsFree

  Print engine launch arguments for copy-paste.

• agent_agentic_serving_lmcache_envFree

  Print LMCache environment variables.

• agent_agentic_serving_show_configFree

  Show current agentic serving configuration.

• agent_costFree

  Show real-time cost breakdown for a fleet by tier.

• agent_deploywrite actionFree

  Provision a heterogeneous agent fleet across all tiers simultaneously.

• agent_langchain_create_langgraphwrite actionFree

  Create a LangGraph workflow.

• agent_langchain_create_pipelinewrite actionFree

  Create an SGLang pipeline.

• agent_langchain_create_workflowwrite actionFree

  Create a LangChain workflow.

• agent_langchain_testFree

  Test connection to LangChain service.

• agent_langgraph_create_workflowwrite actionFree

  Create a LangGraph workflow.

• agent_langgraph_deploywrite actionFree

  Deploy a workflow.

• agent_langgraph_statusFree

  Get workflow status.

• agent_langgraph_testFree

  Test connection to LangGraph service.

• agent_letta_chatwrite actionFree

  Send a message to a Letta agent.

• agent_letta_createwrite actionFree

  Create a new stateful Letta agent.

• agent_letta_deletewrite actionFree

  Delete a Letta agent.

• agent_letta_listFree

  List Letta agents.

• agent_letta_rememberFree

  Teach a Letta agent a durable fact.

• agent_letta_statusFree

  Show the state of a Letta agent.

• agent_listFree

  List all known agent fleets.

• agent_mem0_addFree

  Store a memory in Mem0 for an agent or user.

• agent_mem0_configureFree

  Configure Mem0 credentials and defaults.

• agent_mem0_deletewrite actionFree

  Delete a memory by ID.

• agent_mem0_forgetwrite actionFree

  Delete all memories matching an entity scope.

• agent_mem0_getFree

  Get a single memory by ID.

• agent_mem0_listFree

  List memories for an entity scope.

• agent_mem0_searchFree

  Search agent/user memories.

• agent_mem0_testFree

  Test connection to Mem0.

• agent_mem0_updatewrite actionFree

  Update a memory by ID.

• agent_planFree

  Plan a heterogeneous agent fleet without provisioning.

• agent_scaleFree

  Scale a single fleet tier up or down without affecting other tiers.

• agent_skill_attachFree

  Attach a skill.md to a Letta agent as a durable memory block.

• agent_skill_initwrite actionFree

  Create a skill.md template for an agent.

• agent_statusFree

  Show live status of a fleet — tier health, KV hit rate, queue depth, cost.

• agent_teardownFree

  Terminate all fleet instances and remove fleet state.

• agent_vector_db_downFree

  Teardown a vector database provisioned for an agent fleet.

• agent_vector_db_upFree

  Provision a vector database for an agent fleet.

• analyticsFree

  Get cost analytics

• checkpoint_deletewrite actionFree

  Delete a checkpoint.

• checkpoint_listFree

  List all checkpoints for a training job.

• checkpoint_promoteFree

  Promote a checkpoint to a final model path for serving.

• checkpoint_restoreFree

  Restore a specific checkpoint for a training job.

• checkpoint_saveFree

  Manually trigger a checkpoint save for a running training job.

• configure_providerFree

  Configure provider credentials

• cost_analyzeFree

  Deep cost analysis of current GPU infrastructure: per-provider breakdown, utilization efficiency, waste identification, and optimization potential.

• cost_optimize_recommendFree

  Generate actionable cost optimization recommendations: spot migration, GPU right-sizing, provider arbitrage, idle shutdown, and density packing.

• cost_simulateFree

  Simulate cost optimization scenarios with ROI projections. Compare current vs optimized infrastructure costs.

• create_postgresql_connectionwrite actionFree

  Create a PostgreSQL database connection with auto-table creation. Returns a connection ID for subsequent operations.

• create_sqlite_connectionwrite actionFree

  Create a SQLite database connection with auto-table creation. Returns a connection ID for subsequent operations.

• database_weaviate_create_collectionwrite actionFree

  Create a Weaviate collection.

• database_weaviate_delete_collectionwrite actionFree

  Delete a Weaviate collection.

• database_weaviate_hybrid_searchFree

  Hybrid vector + BM25 search in a Weaviate collection.

• database_weaviate_insertwrite actionFree

  Insert objects into a Weaviate collection.

• database_weaviate_list_collectionsFree

  List Weaviate collections.

• database_weaviate_queryFree

  Vector similarity search in a Weaviate collection.

• database_weaviate_upFree

  Initialize a Weaviate connection.

• deepeval_evaluateFree

  Evaluate a single LLM output with a DeepEval metric (AnswerRelevancy, Faithfulness, Hallucination, etc.).

• deepeval_initFree

  Generate a starter DeepEval test file.

• deepeval_metricsFree

  List available DeepEval metrics for LLM evaluation.

• deepeval_runwrite actionFree

  Run a DeepEval test suite from a Python test file.

• dvc_diffFree

  Show DVC diff between two revisions (e.g. training checkpoints). Shows added, modified, deleted files.

• dvc_pushFree

  Push DVC-tracked data to the configured remote storage.

• dvc_stage_checkpointFree

  Atomic checkpoint staging: DVC add + push + git commit in one operation. Promotes a training checkpoint to versioned storage.

• dvc_statusFree

  Get DVC repository status: tracked files, remotes, and changes since last commit.

• egress_cheapest_routeFree

  Find the cheapest egress route between cloud providers/regions for model weights or dataset transfer. Supports multi-hop routing.

• egress_optimize_stagingFree

  Optimize dataset or model staging across regions by finding the cheapest transfer plan. Integrates with the dataset stager for parallel uploads.

• get_database_connectionFree

  Get information about a database connection including type, status, and configuration.

• governance_compliance_reportFree

  Generate comprehensive compliance report: consent stats, policy evaluations, data movements, violations. For GDPR/SOC2/HIPAA audits.

• governance_evaluate_opaFree

  Evaluate OPA (Open Policy Agent) policies for data access. Checks region restrictions, classification rules, and compliance requirements.

• governance_move_dataFree

  Move data with full governance audit trail. Requires prior consent and OPA policy approval. Tracks integrity, encryption, and compliance.

• governance_movement_historyFree

  Get data movement audit log. Filter by user, dataset, or time range.

• governance_record_consentFree

  Record a consent response (granted or denied) for a pending consent request.

• governance_request_consentFree

  Request user consent for data movement across cloud regions. GDPR/SOC2 compliant consent tracking with audit trail.

• gpu_topologyFree

  GPU NUMA topology report with intra-GPU XCD (Accelerated Compute Die) awareness. Models MI300X (8 XCDs, 192GB HBM3), MI300A (6 XCDs, 128GB), H200 (unified 141GB HBM3e), H100 (80GB). Reports PCIe locality (PIX/PXB/PHB/SYS), GPU-NIC pairing, SR-IOV VF status, and generates XCD-aware NCCL/AITER environment variables.

• guardrails_chatwrite actionFree

  Send a message through NeMo Guardrails and return the safety-filtered response. Applies topical, jailbreak, PII, and factcheck rails.

• guardrails_generate_configFree

  Generate default Colang 2.x guardrails configuration files (topical, jailbreak, PII, factcheck rails).

• guardrails_k8sFree

  Generate Kubernetes deployment manifest for NeMo Guardrails server (standalone or sidecar mode).

• guardrails_testFree

  Test connection to NeMo Guardrails server.

• helm_generateFree

  Generate Helm charts from workload specifications.

• hf_create_endpointwrite actionFree

  Create a HuggingFace Inference Endpoint (paid GPU endpoint). Supports custom GPU types, regions, and scaling.

• hf_delete_endpointwrite actionFree

  Delete a HuggingFace Inference Endpoint.

• hf_endpoint_inferwrite actionFree

  Run inference on a HuggingFace Inference Endpoint. Supports text generation, embeddings, and custom inputs.

• hf_endpoint_infoFree

  Get detailed info about a specific HuggingFace Inference Endpoint: status, URL, scaling config, cost.

• hf_hardware_compareFree

  Compare all hardware options for a HuggingFace model. Returns side-by-side cost, performance, and compatibility analysis.

• hf_hardware_recommendFree

  Get hardware recommendation with cost breakdown for any HuggingFace model. Returns optimal GPU type, estimated cost, and performance score.

• hf_list_datasetsFree

  Search and browse HuggingFace Hub datasets. Filter by author and search query.

• hf_list_endpointsFree

  List all active HuggingFace Inference Endpoints with status, URL, and cost.

• hf_list_modelsFree

  Search and browse HuggingFace Hub models. Filter by author, task, library. Returns model ID, downloads, likes, and tags.

• hf_model_infoFree

  Get detailed model info: architecture, size, downloads, license, tags, pipeline_tag, and model card.

• hf_smart_templateFree

  Auto-generate an optimized deployment template for any HuggingFace model. Analyzes model size, architecture, and quantization to select optimal hardware and generate ready-to-deploy configs.

• hf_space_deploywrite actionFree

  Deploy model to HuggingFace Spaces

• hf_space_statusFree

  Get HuggingFace Space deployment status.

• infer_failoverwrite actionFree

  Run health checks and auto-failover for inference endpoints. If a primary endpoint is unhealthy and has a backup configured, traffic automatically shifts to the backup provider.

• infer_routeFree

  Semantic-aware inference routing. Analyzes query content across 6 signal dimensions (modality, complexity, domain, language, safety, keywords), applies NUMA-aware endpoint scoring, and selects the optimal inference endpoint. Uses DAG parallel execution for signal extraction.

• infer_route_disaggFree

  Disaggregated Prefill/Decode routing (DistServe architecture). Splits LLM inference into compute-bound prefill phase (routed to FLOPS-optimized GPUs like H100 SXM) and memory-bound decode phase (routed to bandwidth-optimized GPUs like MI300X). Tracks KV cache handoffs between endpoint pairs.

• inferx_configureFree

  Configure InferX serverless platform credentials.

• inferx_deletewrite actionFree

  Delete an InferX model deployment.

• inferx_deploywrite actionFree

  Deploy model to InferX serverless platform

• inferx_listFree

  List deployed InferX models

• inferx_optimizeFree

  Get cost analysis for inference endpoints

• inferx_quoteFree

  Get InferX pricing quotes for a GPU type.

• inferx_statusFree

  Check InferX endpoint status

• inferx_usageFree

  Get InferX account usage statistics: requests, cost, GPU hours, latency.

• k8s_createwrite actionFree

  Create Kubernetes cluster with GPU nodes for optimal multi-cloud deployment

• k8s_destroywrite actionFree

  Destroy a Kubernetes cluster

• k8s_device_pluginFree

  Configure Kubernetes GPU device plugin settings: time-slicing, MIG strategy, and resource naming.

• k8s_gpu_operator_installFree

  Install NVIDIA GPU Operator on a Kubernetes cluster. Configures driver containers, device plugin, DCGM exporter, and GPU Feature Discovery.

• k8s_infoFree

  Get information about a specific cluster

• k8s_listFree

  List Kubernetes clusters

• k8s_mig_configureFree

  Configure Multi-Instance GPU (MIG) partitioning on A100/H100 GPUs. Splits a single GPU into isolated instances for multi-tenant workloads.

• k8s_time_slicingFree

  Configure GPU time-slicing for Kubernetes. Allows multiple pods to share a single GPU with configurable oversubscription.

• kserve_generate_yamlFree

  Generate a GPU-aware KServe InferenceService YAML manifest with NUMA pinning, resource limits derived from model size and VRAM, and topology hints.

• kserve_listFree

  List KServe InferenceServices in a Kubernetes namespace.

• kserve_statusFree

  Get detailed status of a KServe InferenceService including readiness, traffic split, and URL.

• langchain_create_sglang_pipelinewrite actionFree

  Create an SGLang model-serving pipeline via LangChain. Connects LangChain agents to SGLang inference endpoints.

• langchain_create_workflowwrite actionFree

  Create a LangChain workflow.

• langfuse_configureFree

  Configure Langfuse credentials (public key, secret key, host URL).

• langfuse_datasetsFree

  List Langfuse datasets for evaluation and fine-tuning.

• langfuse_export_training_dataFree

  Export Langfuse traces as instruction/response pairs for LoRA fine-tuning. Filters by quality score.

• langfuse_k8sFree

  Generate Kubernetes deployment manifest for self-hosted Langfuse.

• langfuse_otel_envFree

  Print OTEL environment variables for instrumenting LLM apps to send traces to Langfuse.

• langfuse_qualityFree

  Get aggregated quality metrics from Langfuse scores for drift detection.

• langfuse_scorewrite actionFree

  Create an evaluation score for a Langfuse trace (e.g. quality, accuracy, relevance).

• langfuse_scoresFree

  List evaluation scores from Langfuse, optionally filtered by trace or score name.

• langfuse_testFree

  Test Langfuse connectivity and list accessible projects.

• langfuse_traceFree

  Get a single Langfuse trace with all observations/spans.

• langfuse_tracesFree

  List recent LLM traces from Langfuse.

• langgraph_create_workflowwrite actionFree

  Create a LangGraph stateful workflow with monitoring. Supports agent graphs, tool calling, and state persistence.

• langgraph_evaluation_workflowwrite actionFree

  Create an evaluator-optimizer workflow in LangGraph. Generates outputs, evaluates quality, and iteratively improves.

• langgraph_orchestrator_workerwrite actionFree

  Create an orchestrator-worker pattern workflow in LangGraph. The orchestrator delegates tasks to specialized worker agents.

• langgraph_workflow_statusFree

  Get the status and metrics of a LangGraph workflow execution.

• local_scanFree

  Scan local machine and network for available GPU devices. Returns total VRAM pool for local-first provisioning.

• lora_addFree

  Hot-load a LoRA adapter onto a running vLLM endpoint. The adapter becomes immediately available as a model name for inference requests. Uses vLLM's fused_moe_lora kernel for 454% higher output tokens/sec on MoE models.

• lora_listFree

  List LoRA adapters loaded on a running vLLM endpoint. Shows base models and hot-loaded fine-tuned adapters.

• lora_removewrite actionFree

  Hot-unload a LoRA adapter from a running vLLM endpoint. Frees GPU memory for other adapters.

• manage_instanceFree

  Manage GPU instances (stop/start/terminate)

• manifestsFree

  List cached manifests and versions for jobs.

• ml_vllm_lora_linkFree

  Load the active registry version of an adapter onto a vLLM server.

• ml_vllm_lora_listFree

  List LoRA adapters currently loaded on a vLLM server.

• ml_vllm_lora_loadFree

  Hot-load a LoRA adapter onto a running vLLM server.

• ml_vllm_lora_syncFree

  Synchronize an adapter from the registry across multiple vLLM replicas.

• ml_vllm_lora_unloadFree

  Hot-unload a LoRA adapter from a running vLLM server.

• mlflow_list_experimentsFree

  List MLflow experiments on the configured tracking server.

• mlflow_log_runwrite actionFree

  Log a Terradev training run to MLflow with auto-injected GPU type, provider, cost/hr, and duration as params.

• mlflow_register_modelFree

  Register a trained model in the MLflow model registry with Terradev provenance tags.

• moe_deploywrite actionFree

  Deploy Mixture-of-Experts models with production-ready cluster templates. Auto-applies vLLM cost optimizations (KV cache offloading for up to 9x throughput, MTP speculative decoding for up to 2.8x speed, sleep mode for 18-200x faster restarts). Supports GLM-5, Qwen 3.5, Mistral Large 3, DeepSeek V4, Llama 5. Configures NVLink topology, tensor parallelism, FP8 quantization, vLLM/SGLang backends, and GPU-aware HPA autoscaling.

• ollama_chatFree

  Chat with an Ollama model using the chat/completions API.

• ollama_generateFree

  Generate text using an Ollama model (non-chat completions).

• ollama_listFree

  List models available on an Ollama server.

• ollama_model_infoFree

  Get detailed information about an Ollama model (parameters, template, license).

• ollama_psFree

  List currently running Ollama models.

• ollama_pullFree

  Pull a model to an Ollama server on a remote instance.

• optimizeFree

  Find cheaper alternatives for running instances

• orchestrator_evictFree

  Evict a model from GPU memory.

• orchestrator_inferFree

  Test inference with a model via the orchestrator.

• orchestrator_loadFree

  Load a model into GPU memory.

• orchestrator_registerFree

  Register a model with the orchestrator.

• orchestrator_startFree

  Start the model orchestrator for multi-model GPU sharing with eviction policies.

• orchestrator_statusFree

  Get orchestrator and model status including GPU memory utilization.

• phoenix_k8sFree

  Generate Kubernetes deployment manifest for self-hosted Arize Phoenix server.

• phoenix_otel_envFree

  Generate OpenTelemetry environment variables for instrumenting serving pods with Phoenix tracing.

• phoenix_projectsFree

  List Phoenix projects (trace namespaces).

• phoenix_snippetFree

  Generate Python instrumentation snippet for adding Phoenix tracing to LLM applications.

• phoenix_spansFree

  List recent spans for a Phoenix project. Supports SpanQuery DSL filters like "span_kind == 'RETRIEVER'" or "status_code == 'ERROR'".

• phoenix_testFree

  Test connection to Arize Phoenix server. Returns collector endpoint and project count.

• phoenix_traceFree

  View full execution tree for a specific trace ID. Shows span hierarchy, latencies, and token counts.

• preflightFree

  Pre-training validation: GPU availability, NCCL, RDMA, drivers across all nodes.

• preflight_gpu_checkFree

  GPU-specific preflight validation: NVIDIA drivers, CUDA version, GPU count, NCCL, NVLink topology, NCU stall-signature profiling, and adversarial config verification (V1-V3).

• preflight_network_checkFree

  Network-specific preflight validation: RDMA availability, InfiniBand status, inter-node bandwidth, latency matrix, firewall rules.

• preflight_reportFree

  Generate full preflight validation report with pass/warn/fail per check. Covers GPU drivers, CUDA, NCCL, RDMA, network, disk, and Docker.

• price_intelFree

  GPU price intelligence with quantitative analytics. Computes delta (rate of change), gamma (acceleration), and annualized realized volatility on GPU spot/on-demand prices across 21+ providers. Identifies cheapest time windows and provider arbitrage opportunities.

• price_spot_riskFree

  Spot instance risk assessment per provider. Returns interruption probability, mean time to interruption, and recommended mitigation.

• price_trendsFree

  Get GPU price trend analysis with delta (rate of change), gamma (acceleration), and annualized volatility. Identifies cheapest time windows.

• provision_gpuFree

  Provision GPU instances for optimal parallel efficiency

• qdrant_collectionsFree

  List all Qdrant vector collections with their point counts and configurations.

• qdrant_countFree

  Count points (vectors) in a Qdrant collection.

• qdrant_create_collectionwrite actionFree

  Create a Qdrant vector collection. Auto-configures vector dimensions from embedding model name.

• qdrant_infoFree

  Get detailed info and stats for a Qdrant collection.

• qdrant_k8sFree

  Generate Kubernetes StatefulSet manifest for self-hosted Qdrant vector database.

• qdrant_testFree

  Test connection to Qdrant vector database. Returns cluster info and collection count.

• query_databasewrite actionFree

  Execute a SELECT query on a database connection. Returns query results as a list of dictionaries.

• ray_disagg_pd_deploywrite actionFree

  Generate a Ray Serve LLM disaggregated Prefill/Decode deployment. Splits inference into compute-bound prefill and memory-bound decode phases with KV cache transfer via NIXL.

• ray_list_jobsFree

  List all running Ray jobs and tasks.

• ray_parallelism_strategyFree

  Compute optimal TP/DP/EP parallelism strategy for a given MoE model and GPU count. Returns recommended configuration with rationale.

• ray_startFree

  Start a Ray cluster (head node or worker). For distributed ML training and inference.

• ray_statusFree

  Get Ray cluster status including node count, resources, memory, and running jobs.

• ray_stopFree

  Stop the Ray cluster on the current node.

• ray_submit_jobwrite actionFree

  Submit a job script to the Ray cluster for distributed execution.

• ray_wide_ep_deploywrite actionFree

  Generate a Ray Serve LLM Wide-EP (Expert Parallel) deployment for MoE models. Returns Python script and config for distributed MoE serving with EPLB and DeepEP.

• rollbackFree

  Explicit versioned rollback. Format: job@version (e.g., llama3@v3).

• run_workflowwrite actionFree

  Run a declarative YAML workflow that chains multiple Terradev commands (provision → preflight → train → monitor → checkpoint). Returns step-by-step execution status with cost estimates and confirmation gates for expensive operations.

• secrets_envFree

  Print environment-style export lines for a provider. By default values are masked.

• secrets_getFree

  Retrieve a stored secret. By default the value is masked.

• secrets_listFree

  List stored provider and key names. Values are never shown.

• secrets_removewrite actionFree

  Remove a provider or a single key from the secret store.

• secrets_runwrite actionFree

  Run a shell command with secrets injected into the environment.

• secrets_setFree

  Store a secret in the Terradev secrets store. Values are encrypted at rest.

• secrets_syncFree

  Import TERRADEV_* environment variables into the secret store for supported cloud providers.

• secrets_verifyFree

  Check which providers are fully configured and which keys are missing.

• setup_providerFree

  Get setup instructions for a provider

• sglangFree

  Complete SGLang optimization stack with workload-specific auto-tuning for 7 workload types: agentic chat, batch inference, low latency, MoE models, PD disaggregated, structured output, and RAG.

• sglang_inferenceFree

  [DEPRECATED] Use sglang action='inference' instead. Test inference against a running SGLang endpoint.

• sglang_startFree

  [DEPRECATED] Use sglang action='start' instead. Start an SGLang inference server on a remote instance.

• sglang_stopFree

  [DEPRECATED] Use sglang action='stop' instead. Stop an SGLang server on a remote instance.

• stageFree

  Compress, chunk, checksum, and position datasets near compute. Supports local paths, S3/GCS URIs, HTTP URLs, and HuggingFace dataset names. Returns staging plan with agent recommendations.

• statusFree

  View all instances and costs with state optimization

• trainFree

  Launch distributed training on provisioned GPU nodes. Supports torchrun, deepspeed, accelerate, and megatron. Use from_provision='latest' to auto-resolve node IPs from your last provision command.

• train_detect_stragglersFree

  Detect straggler nodes in distributed training. Identifies GPUs with significantly lower utilization that slow the whole job.

• train_monitorFree

  Real-time GPU monitoring for training jobs. Shows utilization, memory, temperature, power, and cost.

• train_resumeFree

  Resume a training job from its latest checkpoint. Rebuilds config with topology revalidation.

• train_snapshotFree

  Get complete training monitoring snapshot: GPU metrics (utilization, memory, temp, power), training metrics (loss, grad_norm, lr, throughput), straggler detection, and cost estimate.

• train_statusFree

  List all training jobs and their state (created, running, completed, failed).

• train_stopFree

  Stop a running training job. Kills training processes on all nodes in parallel.

• train_unsloth_runwrite actionFree

  Run an Unsloth local model server.

• train_unsloth_startFree

  Start a coding agent backed by Unsloth's local model server.

• train_unsloth_stopFree

  Stop a running Unsloth server started with `unsloth run`.

• training_config_generateFree

  Generate a complete training configuration from a declarative spec. Auto-detects framework, sets optimal parallelism, and configures distributed training.

• training_launch_distributedFree

  Full distributed training launch with framework auto-detection, topology validation, and monitoring. Combines preflight + train + monitor in one operation.

• upFree

  CLI-native provisioning with manifest cache and drift detection. Use --fix-drift to detect and auto-fix drifted infrastructure.

• upsert_databasewrite actionFree

  Insert or update data in a database table. Performs upsert operation (insert or update on conflict).

• vllm_analyzeFree

  Analyze a running vLLM server's workload and return specific optimization recommendations with before/after comparisons.

• vllm_auto_optimizeFree

  Automatically optimize vLLM configuration by analyzing workload patterns. Selects optimal settings for the 6 critical knobs based on live endpoint metrics or sample request files.

• vllm_benchmarkFree

  Benchmark a vLLM endpoint with concurrent requests. Returns throughput (req/s), success rate, and total latency.

• vllm_inferenceFree

  Test inference against a running vLLM endpoint (completions or chat).

• vllm_infoFree

  Get vLLM server info: loaded models, config, and health status.

• vllm_sleepwrite actionFree

  Put a vLLM server to sleep. Level 1: offload to CPU (fast wake). Level 2: discard weights (minimal RAM).

• vllm_startFree

  Start a vLLM inference server on a remote instance via SSH/systemd. Supports Multi-LoRA, Sleep Mode, KV Offloading, Speculative Decoding.

• vllm_stopFree

  Stop a vLLM server on a remote instance.

• vllm_wakeFree

  Wake a sleeping vLLM server. For Level 2 sleep, also reloads weights and resets prefix cache.

• wandb_create_dashboardwrite actionFree

  Create a custom W&B dashboard with GPU metrics, training loss, and cost panels.

• wandb_create_reportwrite actionFree

  Create a W&B report with custom sections, charts, and narrative text.

• wandb_create_terradev_alertswrite actionFree

  Auto-create standard Terradev alerts: GPU cost > budget, utilization < 50%, training loss spike, straggler detection.

• wandb_create_terradev_dashboardwrite actionFree

  Auto-create a Terradev-specific W&B dashboard with GPU utilization, cost tracking, training metrics, and infrastructure panels.

• wandb_create_terradev_reportwrite actionFree

  Auto-generate a Terradev infrastructure report: GPU costs, provider comparison, training efficiency, and recommendations.

• wandb_dashboard_statusFree

  Get comprehensive W&B monitoring overview: dashboards, reports, alerts, active runs.

• wandb_list_projectsFree

  List all Weights & Biases projects for the configured entity.

• wandb_list_runsFree

  List runs in a W&B project with status, metrics summary, and config.

• wandb_run_detailswrite actionFree

  Get detailed info, metrics, and artifacts for a specific W&B run.

• wandb_setup_alertsFree

  Set up custom W&B alerts for GPU metrics: cost thresholds, utilization drops, training anomalies.

• warm_pool_startFree

  Start the warm pool manager for intelligent model pre-warming. 5 strategies: traffic_based, time_based, priority_based, cost_optimized, latency_optimized.

• warm_pool_statusFree

  Get warm pool status: hit rate, cold starts, memory saved, cost saved.

• Code scan297 source files scanned15/25
• Live reliabilityremote reachable in 183ms20/20
• Tool poisoning245 tool descriptions checked15/15
• Auth qualityopen endpoint exposes 56 write-action tools with no auth3/15
• Maintenancerepository not readable: repo not found3/15
• Maintainer identityno repository or website to verify2/10

Grade history

• 2026-09-19restoreD → Cscore 58: Write-action tools reachable without authentication; eval / new Function used; subprocess with shell=True

Terradev: common questions

Is Terradev MCP server safe?

With care: it is graded C, so read the findings first (58/100). Read the Terradev safety report

How do I install Terradev?

It runs remotely at terradev-mcp.terradev.cloud. Add it to Claude Code, Claude Desktop or Cursor with the snippets above, or call it through the mcp.market gateway without installing anything.

Does Terradev need an API key?

Yes. The registry entry asks for RUNPOD_API_KEY, VAST_API_KEY, LAMBDA_API_KEY, HF_TOKEN and 3 more.

Is Terradev maintained?

The latest release is v6.2.20.

Is Terradev up?

100% of our last 5 checks got an answer. We check remote servers about four times a day.