Skip to main content

Agent Modes

Toporic has three agent modes, each designed for a different kind of work. Choose the one that fits your task.

ModeBest forHas file tools?Has web tools?
CodingSoftware development, debugging, refactoringYesYes
AssistantGeneral questions, writing, brainstormingNoNo
ResearchLooking up documentation, APIs, newsNoYes

Choosing a mode

# Coding mode (default)
toporic

# Assistant mode
toporic --mode assistant

# Research mode
toporic --mode research

You can also switch modes at runtime in the TUI via Ctrl+, (config overlay) or by typing /mode assistant in the input box.

Coding mode

The full toolset. The agent can read and edit your files, run shell commands, search your codebase, and fetch information from the web. This is the default mode — if you don’t specify one, Toporic starts in coding mode.

Use coding mode when you want the agent to do things in your project: write code, fix bugs, run tests, refactor modules, install dependencies, or check the web for documentation.

Add a rate limiter middleware to the API.
Find all usages of the deprecated function and replace them.
Run the test suite and fix any failing tests.

Assistant mode

Pure conversation with no tool access. The agent cannot read or edit files, run commands, or access the web. Think of it as a knowledgeable chat partner embedded in your terminal.

Use assistant mode for questions, explanations, writing help, brainstorming, or anything that doesn’t involve your project files.

Explain the difference between REST and GraphQL.
Draft a README outline for my new library.
What are the trade-offs between async and threaded servers?

Research mode

Web tools only — no file access. The agent can search the web, fetch page content, and call REST APIs, but it cannot touch your project files.

Use research mode to gather information: look up documentation, compare libraries, check API specs, or stay current on a topic.

Research the latest best practices for Rust error handling.
Compare the pricing of three cloud providers for a small API.
Find the GitHub API endpoint for listing repository issues.

Permission modes

Within each agent mode, you also choose a permission mode that controls how much the agent can do without asking:

ModeEffectHow to enable
AutoAll actions run without promptsDefault. TUI: press Tab until [AUTO] shows
BuildMutating actions ask for approval--goal --build "..." or /build in TUI
PlanOnly shows a plan, executes nothing--goal --plan "..." or /plan in TUI

See Tools & Permissions for detailed permission configuration.