Introducing MenteE SWE: An Autonomous Coding Agent for Your Terminal

MenteE SWE is our autonomous software-engineering agent. It lives in your terminal, investigates your repository, plans the change, edits files with surgical precision, verifies with your own tests, and reports back with evidence — never claiming success without proof.

What it is

MenteE SWE is a model-agnostic autonomous SWE agent. You bring your own model key; MenteE supplies the brain-to-editor harness: the tool layer, the verification loop, context management, and orchestration. The product is the harness — not the underlying LLM.

• Autonomous by default — describe a task in plain English; it explores, edits, and verifies on its own.
• Model-agnostic — switch providers without changing how you work.
• Safe by design — no silent deletions, no unbounded reads, no destructive commands without approval.
• Persistent — remembers prior conversations and project knowledge across restarts.

Install and go

npm i -g @menteeai/menteeswe
mentee config
cd your-project
mentee "fix the failing auth tests"

Or try without installing: npx @menteeai/menteeswe "your task". Requires Node 20+. MIT licensed.

The tool suite

MenteE SWE ships with a complete tool layer: filesystem (read, write, patch, move), search (regex, files, list), git (status, diff, log, branches), execution (with safe auto-approve for reads), testing (run tests, linter, typecheck), environment inspection, persistent memory, and web search. There is no delete_file tool — deletion is never automatic.

Supported providers

Z.ai Coding is the default. v0.3.5 also ships 12 more — OpenAI, Anthropic, Gemini, DeepSeek, Kimi, GLM, Qwen (international + China), OpenRouter, and NVIDIA NIM — all configurable via mentee config or environment variables.

What's next

Subagents for parallel investigation, PR creation workflows, and team/shared memory scopes are on the roadmap. Install it today from npm or visit /products/swe for details.