22 tools from the Octomil MCP Server, categorised by risk level.
View the Octomil policy →analyze_files Read multiple files and answer a question about them. benchmark_model Benchmark inference engines for a specific model, measuring tokens/second and latency. compress_prompt Compress a prompt to reduce token count before inference, saving cost without losing meaning. detect_engines Detect which inference engines are available on this machine. detect_hardware Detect full hardware capabilities including CPU, GPU, RAM, and recommended inference backend. embed_text Generate text embeddings using the Octomil platform. Requires OCTOMIL_API_KEY. explain_code Explain what code does in plain English. 2/5 get_metrics Get current model, engine, hardware, and device status. list_models List all available models in the Octomil catalog with publisher, parameter count, engines, and variants. optimize_model Optimize a model for on-device deployment via server-side pruning and quantization. Requires OCTOMIL_API_KEY. recommend_model Recommend the best model configuration for this hardware based on available RAM, GPU, and CPU. review_code Review code for bugs, security issues, and improvements. 2/5 review_file Read a file from disk and review it locally using the on-device model. 2/5 scan_codebase Scan a codebase to find all ML inference points across iOS, Android, and Python code. 2/5 convert_model Convert a local PyTorch model to edge-optimized formats (ONNX, CoreML, TFLite). 2/5 generate_code Generate code from a natural language description. 2/5 plan_deployment Dry-run a deployment to see the plan without executing it. Requires OCTOMIL_API_KEY. 2/5 resolve_model Resolve a model name to engine-specific artifacts including HuggingFace repo, filename, engine, and architecture. 2/5 write_tests Generate unit tests for the given code. 3/5 The Octomil MCP server exposes 22 tools across 3 categories: Read, Write, Execute.
Use Intercept, the open-source MCP proxy. Write YAML rules for each tool — rate limits, argument validation, or deny rules — then run Intercept in front of the Octomil server.
Octomil tools are categorised as Read (14), Write (5), Execute (3). Each category has a recommended default policy.
Open source. One binary. Zero dependencies.
npx -y @policylayer/intercept