Gyaanmitra AI

Tools, Rules & Context

What Gyaanmitra AI can do in your workspace, how it decides what to read, and how project rules steer it.

Built-in tools

Gyaanmitra AI's main abilities are separate tools. It picks among them based on what you asked:

Read & search files

Reads files you point at, and searches the workspace by filename or content to find relevant code on its own.

Edit files

Proposes a change as a diff against the current file. Nothing is written until you approve it.

Run terminal commands

Runs shell commands in the integrated terminal: installing a dependency, running tests, checking git status.

Drive the Integrated Browser

Opens a page, clicks, types, and reads back the state of a running web app, so you can see whether a change actually works.

Approvals

Gyaanmitra AI asks before it does something that is not safely reversible. File edits always show up as a diff first. Terminal commands are checked against a policy you can change. Read-only or clearly low-risk commands (such as git status or ls) can run without a pause. Anything that writes, deletes, or hits the network asks, by default, for a one-time approval, or an "always allow this command" approval. Tune this in Gyaanmitra AI Settings → Terminal.

Context: what Gyaanmitra AI sees

By default, context includes the file you have open and a summary of the workspace layout. You can add more, or keep it smaller:

@ mentions: type @ in the chat input, then a filename, folder, or symbol, to include it on purpose.
Add Files: the paperclip-style control next to the chat input, for attaching a file without typing its name.
Context usage indicator: shown near the input, so you can see roughly how much of the model's context window this conversation is using.

How rules steer behavior

Rules are plain-language instructions that apply on their own, so you do not repeat yourself every conversation. Examples: "always write tests alongside new functions," "use the project's existing error-handling pattern," "never touch the generated/ folder." Manage them from Gyaanmitra AI Settings → Rules. They can apply everywhere, or only in one workspace.

Skills, Plugins & Subagents

1

Skills

Named playbooks for a kind of task you do often. A skill can bundle instructions, example code, and the tools you prefer, so Gyaanmitra AI follows a consistent approach when that skill is relevant.

2

Plugins

Add what Gyaanmitra AI can reach by connecting Model Context Protocol (MCP) servers or other integrations under Gyaanmitra AI Settings → Tools & MCP. That gives it access to systems beyond your local workspace, when you choose to.

3

Subagents

Hand a self-contained piece of work to a subagent, such as reading an unfamiliar part of the codebase or running a long test suite. It works on its own and reports back, so your main conversation stays focused.

Hooks

Hooks run a script of yours at set points in Gyaanmitra AI's workflow: before a file is edited, after a command finishes, when a conversation starts. Use them to keep a team habit (format every file it touches) or to log what it did so you can review it. Configure them under Gyaanmitra AI Settings → Hooks.