BloxAgent - Custom AI Exploit Dev [MOBILE SUPPORT]

Agentic AI In-Game: Runs a full AI agent inside any live Roblox game. Give it a task in plain English and it carries it out — reading the game, writing and running Luau, editing scripts, and reporting back when done.

Multi-Provider Support: Bring your own model. Works with Ollama, OpenAI, Anthropic, DeepSeek, Moonshot, Venice AI, and Zhipu GLM out of the box. Switch providers and models from a dropdown, or fetch the live model list straight from your provider's API.

Code Execution: The agent writes and runs real Luau on demand and reads back prints, return values, warnings, and errors, with full executor access.

Background Processes: Run continuous scripts without blocking the chat. Get a process ID back, then list, read output, and kill them individually or all at once.

Game Inspection: Pulls PlaceId, GameId, server JobId, your character state, and the workspace tree so the agent orients itself before acting.

Instance Finder: Search the whole game by name and get back full paths, class names, and positions for parts and models.

Script Codebase Tools: List every Script, LocalScript, and ModuleScript with full paths. Read source where the decompiler allows, edit with an automatic line diff, and revert the last edit from a snapshot.

Subagents: Fan big jobs out across multiple independent subagents running in parallel, then collect each result by ID.

Built-In Executor: A clean, Solara-style script editor in its own draggable window. Multiple tabs, line-number gutter, Lua syntax highlighting, run button, and autosave so scripts survive between sessions.

Agent-Executor Collaboration: The AI can see your open tabs, read what you're working on, rewrite a tab with a live diff, open new tabs with starter code, and run any tab to test it.