Question 1

What is AUREM Dev?

Accepted Answer

AUREM Dev is an autonomous AI engineer. You connect a GitHub repository, describe what you need in chat, and AUREM plans, writes, tests and ships the change — committing directly to your branch via the GitHub REST API in minutes. If your team requires review, you can switch any project to PR-based delivery instead and AUREM will open a real Pull Request with the same atomic diff.

Question 2

How is AUREM Dev different from Cursor, Devin or Replit Agent?

Accepted Answer

Cursor is an IDE assistant; you still write the code. Devin and Replit Agent run in their own sandboxes that you have to copy code back from. AUREM Dev is repository-first — it grounds itself in your existing files, respects your layout and conventions, and lands changes straight on your branch via the GitHub API instead of asking you to copy snippets. You skip the review-then-merge round-trip (or keep it, your choice).

Question 3

What developer tools (skills) does AUREM Dev have?

Accepted Answer

AUREM Dev ships with 22 native developer skills grouped into five categories: READING (semantic_search_repo, read_repo_file, read_repo_files, list_repo_files, search_repo), INTEL (find_usages, get_dependencies, get_env_vars, detect_framework, get_repo_info), GITHUB (get_commit_history, get_commit_diff, list_issues, get_pr_comments), WEB (web_search, web_search_and_summarize, fetch_url, firecrawl_scrape, firecrawl_crawl_site, find_package_docs) and VALIDATE (validate_syntax, e2b_run_code). The agent picks the right combination automatically — for example, before refactoring a function it calls find_usages to enumerate every caller. This is the same industry-ceiling tool count Claude Code and Cursor use.

Question 4

Is there a free tier?

Accepted Answer

Yes. Every new account gets 10 free tasks per month forever — no credit card required. Paid plans are flat fee: Starter $9/mo (50 tasks, Swift mode), Pro $19/mo (300 tasks, Swift + Pro review modes + automations), Team $49/user/mo (400 tasks/user, Swift + Pro + Maxx with Claude Sonnet 4.5 watchdog, priority queue + admin panel). Annual plans save 20% (e.g. Pro $182/year vs $228 monthly). There is no token billing and no credit pool to drain.

Question 5

Which AI models does AUREM Dev support?

Accepted Answer

AUREM Dev is model-agnostic. You can bring your own API key for Anthropic (Claude Sonnet 4.5), DeepSeek V3, OpenAI (GPT-5.2), or Google (Gemini 3 Flash / Gemini 3 Pro). The platform also ships with a managed Universal Key that works out of the box across all four providers — no setup needed.

Question 6

Does AUREM Dev have access to my private code?

Accepted Answer

Only when you explicitly connect a repository with a GitHub Personal Access Token. Tokens are encrypted using HKDF-Fernet and stored per-tenant in our vault. Only the files needed for the current task are loaded into the model context. You can revoke access at any time from Settings, and the encrypted token is wiped from our vault on disconnect.

Question 7

How does AUREM Dev push changes to GitHub?

Accepted Answer

Two delivery modes, you choose per project. Default mode: AUREM commits straight to your configured branch (main by default) via the GitHub REST API — trees + blobs + commits + refs — producing one atomic commit per task. No local git binary required, no PR round-trip. Optional PR mode: AUREM opens a real Pull Request via the github_deploy_service flow if your team requires review. Both modes pass through the Vanguard 007 security scanner (25+ patterns) and the Claude Sonnet 4.5 Vanguard Verify Agent before any byte lands.

Question 8

Is AUREM Dev secure for production codebases?

Accepted Answer

Three layers of pre-commit gates: (1) Vanguard 007 AST scanner catches 25+ insecure patterns — hardcoded secrets, eval(), shell injection, unsanitized template strings, vulnerable npm/pip packages. (2) Vanguard Verify Agent re-reads the proposed diff with Claude Sonnet 4.5 and blocks anything that looks dangerous. (3) Snippet-level syntax + e2b sandbox validation before any line lands in the PR. All three log to a per-tenant audit table you can inspect at /admin/vanguard.