v1.0.25 · standalone binary · no node required

the terminal is your AI IDE

Netro is a terminal-native AI coding agent. It reads, writes, and runs your code — with workflows, subagents, persistent memory, installable skills, and a live HUD. One binary. Every platform.

$ curl -fsSL https://www.netrocli.bond/install.sh | bash
PS> irm https://www.netrocli.bond/install.ps1 | iex

or scroll down to try the live demo terminal ↓

netro — my-project demo
netro · ctx 0% · ---------- · cost $0.000 · claude-sonnet-4-5

## what's inside — tab through the panes

Everything you'd expect from a coding agent — plus the things you didn't know you needed.

 Read(src/server.ts)
  ⎿  Read 214 lines
 Edit(src/server.ts)
    - app.listen(3000)
     + app.listen(process.env.PORT ?? 3000)
 Bash(npm test)
    ✓ 34 passed (1.4s)
 Done. Port now reads process.env.PORT, falls back to 3000.

Tools: read, bash, edit, write, grep, find, ls. The agent plans, executes, and verifies — all in your terminal.

 /goal refactor auth to JWT
  → spec phase: analyzing codebase…
  → design phase: 3 files to change, 2 new
  → implement phase: editing src/auth.ts…
  → review phase: running tests… ✓
  → done · 4 files changed · all tests pass

workflows: /goal · /workflow · /feature-dev
           /code-review · /security-review

Multi-phase workflows run autonomously — spec → design → implement → review → deploy. Or drive them manually step by step.

 498 skills loaded
   git-review         core
   deploy-check       devops
   spec-writer        planning
   system-prompt-design agent
   tdd-workflow       testing
  …and 493 more — install from GitHub

/skills install github.com/you/repo

Skills are SKILL.md files that teach netro new procedures. They hot-reload on change. Install from any GitHub repo.

 dispatching parallel subagents…
  explorer  → scanning src/ for patterns
  architect → designing the new API
  reviewer  → checking edge cases
  scout     → searching for prior art

   all reports merged · 3 findings
  → implementing based on agent synthesis…

Parallel and chained subagents — each with isolated context. Partition by file or module to avoid conflicts.

netro · claude-sonnet-4-5 · ctx 42%
┌─────────────────────────────────────────┐
 ▮▮▮▮▮     ▯▯▯▯▯ 42% context   
 tools: read ✓ edit ✓ bash ✓       
 todos: 3/5 done                    
 cost: $0.031                        
└─────────────────────────────────────────┘

Live statusline: context usage, tool calls, todo progress, session cost. Claude-style tool display and diff rendering.

 loading persistent memory…
   user prefers Indonesian responses
   project uses erasable TS syntax only
   no enum, no param properties
   last session: fixed auth bug in server.ts

~/.netro/agent/
├── memory/      # durable facts across sessions
├── skills/      # hot-reloaded procedures
├── agents/      # subagent definitions
└── sessions/    # full session history (jsonl)

Memory persists across sessions. Skills grow permanently. Every session is saved — fork, clone, or tree-navigate.

## everything is a file

No database, no cloud. Your agent's entire state lives in ~/.netro/agent/ — JSON, markdown, and jsonl. Inspect it, version it, delete it.

settings.json

      

## pricing

Free with your own API keys. Or subscribe for hosted model access — no keys needed.

Free
$0
  • All agent features
  • Bring your own API keys
  • 498+ installable skills
  • Subagents & workflows
  • Persistent memory
Get started
Enterprise
Talk
  • Everything in Lifetime
  • Private model endpoints
  • Custom skills & agents
  • Priority support
  • Deployment assistance
Contact us

Free forever with your own keys. No credit card. How subscriptions work →

Ready to code at the speed of thought?

$ curl -fsSL https://www.netrocli.bond/install.sh | bash