Auggie CLI
Augment's agentic coding CLI for your terminal: it understands your codebase and ships changes via natural language — analyzing code, making safe edits, automating routine tasks.
- Source repo
- augmentcode/auggie
- Stars
- ★ 289
- Last updated
- 1d ago
- License
- NOASSERTION
- Primary language
- Shell
- FA score
- 52/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specific
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer working in the terminal who wants an agent to understand the whole codebase and make changes from natural-language instructions
- Not a fit if
- Teams that need fully local operation with code never leaving the machine
- Users who don't want to depend on Augment's cloud service and login
- Source review
- 52/100 · Major gaps
What does this agent do, and when should you use it?
Auggie is Augment's official agentic coding CLI that runs in your terminal, distributed as the npm package @augmentcode/auggie and requiring Node.js 22 or later. It runs inside your project directory, where it analyzes code, makes safe edits, and automates routine tasks from natural-language instructions. You must run auggie login to authenticate with an Augment account, so its intelligence depends on Augment's cloud service. Beyond interactive use, it supports a one-shot --print mode for CI and a --quiet flag for final-output-only results. The project also ships a custom slash-command mechanism (Markdown files under .augment/commands/) and official companion GitHub Actions — augment-agent, review-pr, and describe-pr — for automating PR workflows. The repository's license is listed as NOASSERTION, so verify licensing terms before adopting.
You run auggie in a project directory (optionally with an initial prompt); Auggie reads and understands the codebase, then analyzes code, makes edits, and automates routine tasks per natural-language instructions. auggie --print "instruction" runs once and prints to stdout; adding --quiet returns only the final output, making it suitable for CI. Reusable prompts live in .augment/commands/ as Markdown files with frontmatter and become slash commands such as /code-review path/to/file. Inside Auggie or the IDE agent, /feedback submits bug reports with helpful context. The companion GitHub Actions automatically review and describe pull requests.
- A developer working in the terminal who wants an agent to understand the whole codebase and make changes from natural-language instructions
- A CI engineer embedding auggie --print into pipeline scripts for one-shot code analysis or fixes
- A team standardizing code review, bug fixing, security review, and performance optimization as reusable slash commands in .augment/commands/
- A repository maintainer using the official GitHub Actions (review-pr, describe-pr) to improve PR quality automatically
- An existing Augment IDE-agent user who wants the same agentic experience in the terminal
How do you install or deploy this agent?
Requires Node.js 22+. Install globally via npm:
npm install -g @augmentcode/auggie@latestThen log in with your Augment account:
auggie loginHow do you use this agent?
Run in your project directory, optionally with an initial prompt:
cd /path/to/your/project
auggie "optional initial prompt"Run once and print to stdout (great for CI):
auggie --print "your instruction"Return only the final output:
auggie --quietCustom slash commands: create .augment/commands/ in your project root and add Markdown files with frontmatter (e.g., code-review.md, bug-fix.md, security-review.md, performance-optimization.md); invoke them like /code-review path/to/file. Use /feedback in-session to report issues.
What are this agent's strengths and limitations?
- One npm command to install and a single auggie login gets you running within minutes — low setup cost
- --print and --quiet modes are designed for CI, enabling non-interactive use in automation pipelines
- The .augment/commands/ Markdown slash-command mechanism lets teams codify review and fix workflows as reusable assets
- Official companion GitHub Actions (augment-agent, review-pr, describe-pr) extend the agent into PR workflows
- Requires an Augment account login and depends on Augment's cloud service; code analysis is not fully local and you cannot plug in your own models
- The repository license is listed as NOASSERTION, so commercial adopters must verify licensing and legal terms themselves (privacy policy at augmentcode.com/legal)
- No pricing is stated in the README, so usage cost must be confirmed with Augment
- Requires Node.js 22+, so older environments need an upgrade first
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Auggie CLI This agent | 52 · Major gaps | CLI | ★ 289 | 1d ago | Shell | — |
| PR Lens | 61 · Some gaps | Agent plugin / skillFreemium | ★ 2.2k | today | TypeScript | Codex |
| Vet Code Change Verifier | 65 · Some gaps | CLIFree + model costs | ★ 511 | 1mo ago | Python | Codex · Claude Code · OpenAI API · Claude API |
| Continuous Claude: Continuous AI Agent for GitHub PRs | 53 · Major gaps | CLIFree + model costs | ★ 1.4k | 28d ago | Shell | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The core CLI (augment.mjs) is a closed, pre-built artifact; its permission model, tool-confirmation behavior, and data flows cannot be audited from source — scored 1. The Warp plugin's permission_request hooks and tests show a confirmation-event protocol exists, but only for that third-party integration. CI workflows declare minimal permissions (contents: read / pull-requests: write) with pinned Bun version, so dependency security earns 2. The describe-pr workflow auto-sends PR content to an external AI service using a session secret, so external_effects is deducted. No rollback/recovery guidance. Attribution is clear (Copyright Augment Computing, Inc.; SECURITY.md provides security contact).
The Warp plugin hooks have systematic tests covering protocol negotiation, JSON escaping, and routing fallbacks — self_consistency 2. But core functionality depends entirely on a closed npm-distributed package plus an active Augment subscription; use rights terminate when the subscription ends, a single point of failure — scored 1. CI scripts emit clear error messages (missing version, missing signing identity), so failure_messages 2.
README clearly targets scenarios (terminal coding agent, CI --print mode, custom slash commands, GitHub Actions) — 2. Capability boundaries are barely described: which operations modify files or require confirmation is not stated in the source material, hence the deduction. Slash commands and hook event types are cleanly separated — trigger_precision 2. Node 22+ requirement and cross-platform binaries (darwin/linux/windows) are well documented — environment_fit 2.
Repository layout is clear (README, LICENSE, SECURITY, examples, plugin_marketplace) — 2. Install notes include Node version and npm command — 2. Naming is consistent (@augmentcode/auggie / auggie) — 2. Example command templates and plugin tests are provided — 2. Known limitations are thin (only a 1.x migration guide; no data/security limits) — 1. LICENSE.md is complete and specific (grant scope, restrictions, disclaimer) — 3. No in-repo CHANGELOG; versioning relies on npm tags — 1. SECURITY.md commits to vulnerability response and supported versions — 2.
--print/--quiet make output consumable by scripts and CI — 2. Codebase understanding is claimed but unverifiable in static review; the 1.x migration guide suggests active evolution — marginal_value 2. Paid subscription is mandatory and commercial embedding is prohibited, making cost/benefit opaque to adopters — 1.
The core agent is a closed binary; key README claims ('understands your codebase', 'safe edits') cannot be traced to source — claim_traceability 1. README, LICENSE, SECURITY, workflows, and plugin tests are mutually consistent with complete external doc/example links — cross_source_corroboration 2. The test file clearly distinguishes mock inputs from real behavior; facts and inference are separable — 2.
- The core CLI is a closed pre-built binary; this is a static source review with no execution, so all safety and capability conclusions are low-confidence.
- Use requires an active Augment subscription; the license terminates when the subscription ends, and commercial embedding requires separate permission.
- GitHub Actions such as describe-pr send PR content to an external AI service; evaluate data flows and required secrets (AUGMENT_SESSION_AUTH) before adoption.
- No in-repo CHANGELOG; version information must come from the npm registry.
- Privacy and data handling are presented only via external links, not detailed in the repository.