Project-aware context
Aurict detects the framework, package manager, structure, and relevant files before work begins.
open-source terminal agent
An AI coding agent built for the terminal.
Aurict reads your project, delegates work to specialist agents, uses typed tools, and asks before sensitive actions. Bring your own model provider and keep the workflow in the shell you already use.
9 specialist agents218+ contextual skills12 built-in providersmacOS · Linux · Windows
What is a terminal agent?
A terminal agent is an AI assistant that works inside a command-line environment. Unlike autocomplete alone, it can inspect a repository, plan a task, call development tools, edit files, run verification commands, and report what changed.
Aurict turns that pattern into a controlled runtime. Specialized agents handle exploration, implementation, review, tests, documentation, security, debugging, performance, and analytics while the coordinator keeps the task and evidence connected.
The terminal stays the control surface: project context remains visible, shell commands pass through classification and approval rules, and you can switch model providers without moving the codebase into a vendor-specific IDE.
terminal-native capabilities
More than a chat prompt attached to a shell.
A useful coding agent needs context, tools, boundaries, and verification—not just model access.
Multi-agent execution
Nine specialist roles separate exploration, coding, review, testing, security, and other concerns.
Explicit command control
Shell commands are classified before execution; dangerous actions stop for direct approval.
MCP and custom tools
Import compatible MCP configuration and extend the runtime with project or user tools and skills.
Provider choice
Use built-in adapters for leading cloud and local model providers through a consistent workflow.
Verification evidence
Changed files, checks, open work, and waivers remain visible in a durable completion record.
how it works
From request to verified change.
Understand
Scan the codebase and identify the files, dependencies, conventions, and risks that shape the task.
Delegate
Route independent parts to the right specialist agents while preserving one coordinated objective.
Act with boundaries
Use typed tools and scoped permissions. Sensitive commands and exceptional writes still require approval.
Verify and report
Exercise the affected path, run relevant checks, and return evidence with any remaining work.
terminal agent use cases
One runtime, several engineering workflows.
Use Aurict for focused tasks or for work that crosses several parts of a repository.
Repository exploration
Map architecture, dependencies, ownership boundaries, and the files most relevant to a change.
→Refactoring
Plan and implement multi-file changes while preserving conventions and verifying behavior.
→Code review
Inspect changes for correctness, security, performance risks, and missing tests.
→Testing and documentation
Generate meaningful tests and update technical documentation from the actual code paths.
→choose with evidence
Compare terminal coding agents by workflow—not slogans.
Evaluate provider support, permission scope, project context, extensibility, verification, operating-system support, and how clearly the tool reports unfinished work.
terminal agent FAQ
Common questions about AI agents in the terminal.
Is Aurict a terminal agent or an IDE extension?
Aurict is a terminal-native runtime. It runs in your existing shell and does not require moving the project into a vendor-specific editor.
Which operating systems does the Aurict terminal agent support?
Aurict provides compiled binaries for macOS, Linux, and Windows x64. The npm package selects the matching platform binary.
Can I choose the AI model provider?
Yes. Aurict uses a bring-your-own-key model with built-in adapters for supported cloud and local providers. You can switch from the terminal.
Can the agent run terminal commands automatically?
Commands pass through classification and permission controls. Safe actions can run within the active policy; dangerous or out-of-scope actions require explicit approval.
Does Aurict support MCP servers?
Aurict can import compatible MCP server configuration. Use /mcp to verify each connection and inspect the tools it exposes.
Put an agent runtime in your terminal.
Install the open-source CLI, choose a provider, and start inside an existing project.