Navin
All documentation

Navin Harness

Navin is an agent harness. The product name on the site is Navin Harness.

A harness is the loop that runs a model with tools, memory, permissions and channels until the job is done. It is not a chat tab in the browser. It is not autocomplete in an editor.

What the harness does

  1. Take a goal (chat, board mission, channel message, heartbeat, slash command).
  2. Call a model (local, BYOK or managed).
  3. Use tools: files, shell, git, web, scrape, MCP, cron, studios.
  4. Keep memory and session context.
  5. Ask you before risky steps (Git brakes, approvals).
  6. Repeat until the slice is done, then stop for review.

What sits on the harness

  • Code - Vision 360, plan, review, security, debug, board autonomy
  • Channels - WhatsApp, Telegram, Slack, email and more
  • Studios - scraping, leads, marketing, SEO, documents, meeting, media
  • Heartbeat - quiet recurring checks
  • Plugins / skills / MCP - extend the same loop

Local-first

The harness runs on your machine (Windows, macOS, Linux). Chats and files stay on disk. You choose local models, your own keys, or optional managed models.

Related