Gunamaya AI
Tools, Rules & Context
What Gunamaya AI is capable of doing to your workspace, how it decides what to read, and how to steer its behavior with project-level configuration.
Built-in tools
Gunamaya AI's core capabilities are exposed as discrete tools it chooses between based on what you've asked for:
Read & search files
Reads specific files you reference and searches across the workspace by filename or content to find relevant code on its own.
Edit files
Proposes changes as a diff against the current file content — nothing is written until you approve it.
Run terminal commands
Executes shell commands in the integrated terminal — installing a dependency, running a test suite, checking a git status.
Drive the Integrated Browser
Navigates, clicks, types, and reads back the state of a running web app to verify a change actually works.
Approvals
Gunamaya AI asks before doing anything that isn't safely reversible. File
edits are always shown as a diff first. Terminal commands are checked
against a configurable policy: read-only or clearly low-risk commands (like
git status or ls) can be allowed to run without interruption, while
anything that writes, deletes, or hits the network by default asks for a
one-time or "always allow this command" approval. Tune this policy in
Gunamaya AI Settings → Terminal.
Context — what Gunamaya AI sees
By default, context includes your currently active file and a summary of your workspace's structure. Expand or narrow that with:
@mentions — type@in the chat input followed by a filename, folder, or symbol to explicitly include it.- Add Files — the paperclip-style control next to the chat input for attaching files without typing their name.
- Context usage indicator — shown near the input, so you always know roughly how much of the model's context window your current conversation is using.
Rules — steering behavior
Rules are plain-language instructions that apply automatically, so you don't
have to repeat yourself every conversation — "always write tests alongside
new functions," "use the project's existing error-handling pattern," "never
touch the generated/ folder." Manage them from Gunamaya AI Settings →
Rules; they can apply globally or be scoped to a specific workspace.
Skills, Plugins & Subagents
Skills
Reusable, named playbooks for a specific kind of task — a skill can bundle instructions, example code, and preferred tool usage for something you do often, so Gunamaya AI follows a consistent approach every time it's relevant.
Plugins
Extend what Gunamaya AI can do by connecting Model Context Protocol (MCP) servers or other integrations under Gunamaya AI Settings → Tools & MCP, giving it access to systems beyond your local workspace when you choose to.
Subagents
Delegate a self-contained piece of work — researching an unfamiliar part of the codebase, running a long test suite — to a subagent that works independently and reports back, keeping your main conversation focused.
Hooks
Hooks run your own script at specific points in Gunamaya AI's workflow — before a file is edited, after a command finishes, when a conversation starts — useful for enforcing team conventions (auto-formatting every file it touches) or logging its activity for your own review. Configure them under Gunamaya AI Settings → Hooks.