panda1847/oracle-mission-control

ORACLE is a mission orchestration tool built for people who run authorized security testing, the kind of testing where you have explicit permission to probe a computer network for weaknesses. Instead of letting an AI decide what happens during a test, ORACLE uses a plain, predictable state machine, called a deterministic planner, to decide the next step. An AI can suggest an action, but it cannot take one outside of what the planner already allows, and if the AI fails, the planner keeps going without it. Before any action runs, ORACLE checks it against a scope you define ahead of time, meaning the exact list of systems you are allowed to test. Actions are also classified by risk, and higher risk actions can require a person to approve them before they happen. Every finding, host, and service discovered during a test is stored in what the project calls an evidence graph, a structured record that tracks where each piece of information came from and how confident the tool is about it, rather than a plain list you would have to sort through by hand. Under the hood, the project combines a Python package for its command line interface and web control panel with a separate Go program that runs the actual test commands in an isolated, timed environment. To try it, you install the Python requirements, run the Go tests, run the Python tests, then start a demo web dashboard on your own computer. Running it against a real target requires you to state the scope directly on the command line. It can also run using a free, locally installed AI model through Ollama, so no external API key or cost is required. This is the free Community Edition, released under the MIT license, with 2 stars so far. Paid Pro and Enterprise editions exist for teams that need cloud based storage, multiple test runners, and formal compliance reports.