Features
Build — Full access agent
Terminal access
Run any command in a sandboxed system shell.
Run any command in a sandboxed system shell.
File system
Read, write, and organize files across your machine.
Read, write, and organize files across your machine.
Git operations
Commit, branch, create PRs, revert changes.
Commit, branch, create PRs, revert changes.
Code intelligence
LSP-powered project-wide analysis.
LSP-powered project-wide analysis.
Plugin ecosystem
Extend capabilities via MCP protocol.
Extend capabilities via MCP protocol.
Image understanding
Drop in screenshots, diagrams, or mockups.
Drop in screenshots, diagrams, or mockups.
Plan — Read-only agent
Safe exploration
Navigate codebases without making any changes.
Navigate codebases without making any changes.
Architecture analysis
Understand project structure and dependencies.
Understand project structure and dependencies.
Code review
Get feedback and suggestions without file edits.
Get feedback and suggestions without file edits.
Change planning
Detailed plans before execution.
Detailed plans before execution.
Browser Use — Web automation agent
Playwright-powered
Full web automation with Playwright engine.
Full web automation with Playwright engine.
Form filling
Automate multi-step form submissions.
Automate multi-step form submissions.
Data scraping
Extract structured data from any page.
Extract structured data from any page.
UI testing
Automated visual and functional testing.
Automated visual and functional testing.
General capabilities
Native desktop
macOS, Windows, Linux with auto-updates.
macOS, Windows, Linux with auto-updates.
Granular permissions
ACL-based access control per resource.
ACL-based access control per resource.
20+ themes
Catppuccin, Nord, Tokyo Night, and more.
Catppuccin, Nord, Tokyo Night, and more.
18+ languages
Multi-language UI support.
Multi-language UI support.
20+ LLM providers
Claude, GPT, Gemini, Groq, and more.
Claude, GPT, Gemini, Groq, and more.
Self-hosted
No cloud dependency for core features.
No cloud dependency for core features.
Frequently asked questions
FBC Agent provides three distinct agent modes. Build mode gives the AI full terminal and file system access to write code, run commands, manage git, and edit files directly. Plan mode is a read-only agent that explores codebases and analyzes architecture without making any changes. Browser Use mode leverages Playwright to automate web browsing, fill forms, scrape data, and run UI tests.
Build has full read/write access to your terminal and file system. Plan is read-only and safe for code exploration and architecture analysis. Browser Use automates web interactions via Playwright — filling forms, scraping data, and testing UIs.
No. FBC Agent runs entirely on your machine. Your code, files, and conversations never leave your computer unless you explicitly configure a cloud-based LLM provider. Even then, prompts go directly to that provider — FBC Agent does not intermediate or store them.
Yes. FBC Agent has granular permission controls including ACL-based policies. You can restrict terminal access, file system paths, network access, and browser automation permissions per agent.