Enables per-profile tool filtering and enforcement for one or more upstream MCP servers, exposing only allowed tools and rejecting blocked calls, with Streamable HTTP serving and observability.
Exposes DeepSeek Harness tools to any MCP-compatible client over streamable HTTP, allowing allowlisted operations such as file read, glob, grep, and web search while preserving the harness's sandbox and approval pipeline.
Exposes a single HTTP endpoint that aggregates tool definitions from GitHub, Kubernetes, and database namespaces, letting clients list available tools and invoke them by name. It provides health-check, tool-discovery, and tool-call routes for local, read-only experimentation without calling hosted models or applying production changes.
Enables building and testing MCP-style tool servers by exposing list_tools and call_tool over stdio or HTTP with local stub tools such as echo, get_time, and fake_crm_lookup.
Publishes an MCP Streamable HTTP JSON-RPC endpoint where clients can initialize sessions, list and call synthetic deployment, runtime, cluster-inspection, and audit-log tools. While serving those decoy tools, it records initialization, tool discovery, argument shapes, and multi-tool sequences as evidence for classifying and investigating agent-like behavior.