Navin capabilities

This page is the product map: what Navin can do, which module owns it, and where the setting lives in the source UI (navin-cli Ctrl+G, or desktop Settings). Install first: Installation. Mental model: Navin AGI.

Config file: ~/.navin/config.json. Same file for navin-cli and the desktop.

How you run it

Surface	Command / URL	Notes
Terminal AGI	`navin-cli`	Product CLI. Open a project folder first
Desktop / workbench	`navin .`	Same engine, GUI
Gateway	`navin gateway`	Long-running host, default WebUI `:8765`
AGI switches	`navin agi status`	All off by default
Settings	Ctrl+G	Providers, models, tools, safety
Mode	Ctrl+T	ask, plan, agent, review, security, debug
Docs on the web	navin.live/en/docs	Same topics as this `docs/` tree

navin-cli keys: interactive · shortcuts

Agent loop

Navin takes a goal and keeps working until it is done or blocked.

MISSION → PLAN → ACT (files, code, shell, git, browser, MCP)
       → VERIFY (tests, lint, security, evidence)
       → CONTINUE / RETRY / REPLAN

Mode	Slash / habit	What it does
Ask	free text, no writes	Understand the repo
Plan	`/blueprint`	Design only, then hand off
Agent	`/forge` `/cruise` `/mission` `/mobile`	Build, test, iterate
Review	`/inspect`	Read-only review + HTML report
Security	`/fortify` and the security family	AppSec audit + hardening choices
Debug	`/debug`	Reproduce, prove root cause, fix plan

Full mode table: Composer modes · slash list: cli/slash-commands · Code commands: navin_dev/en/commands

Loop and Heartbeat

Loop is a multi-cycle run toward a goal. Examples: /cruise, /mission, studio desks (Career, Trading, Tenders, Marketing, Leads).

Heartbeat is a quiet periodic pass. It reads .navin/HEARTBEAT.md in the project and only reports when something is useful. It is not a noisy cron.

navin gateway --background
navin gateway install-service

Create a scheduled job from chat ("every weekday at 9, …") or edit HEARTBEAT.md.

AGI (off by default)

These are the learning layers. A locked switch unlocks itself when the stage below has passed its exam. Nothing is fine-tuned on your chat model.

Layer	Switch	What it does	Docs
Skills evolution	`navin agi on`	Draft, exam, promote or rollback a skill after repeated failure	skills-evolution
Memory	`navin agi memory on`	Episodes and recall across sessions	memory
World model	`navin agi world on`	Predict what a tool will answer before calling it	world-model
Policy	`navin agi policy on`	Learn the next action from eval trajectories	policy
Transfer protocol	`navin agi transfer status`	Hidden exam + dossier; claim stays forbidden until proofs pass	transfer-protocol

navin agi status
navin agi drafts
navin agi world exam
navin agi policy exam

Flags live under the project: .navin/skills-evolve.json, .navin/cognition.json, .navin/world-model.json, .navin/policy.json.

CLI reference: navin agi

Code studio (`#/code`)

Workbench: explorer, editor, terminals, web preview, mobile preview, git, graph, multi-agent chat.

Piece	What you get	Docs
Workbench	Explorer, editor, terminals, preview, project picker	workbench
Graph	Files / packages, impact, path	graph
Project Home	Resume, tasks, issues, Vision 360, brain, timeline	project-home
Board autonomy	Consent, per-task branch, PR on done, issue sync	board-autonomy
Actions	One-click quality, security, performance, design audits	actions
Expert tools	`code_review`, `security_scan`, `debug_repair`, DebugMCP	expert-tools
Mobile	Expo / RN / Flutter run + preview	mobile · mobile.md
Editor AI	Tab completions, inline edit, diff review	editor-ai
Plugins / skills	Packs and project skills	plugins · skills
App templates	Install a full app from the gallery	app-templates
Evolve	Prove a change under load before merge	navin_evolve
RiskLens	Assume failure in 6 months, revise first	navin_risklens

Index: navin_dev

Other studios

Same agent, same memory, different desk.

Studio	Route	Typical slash	Docs
Scraping	`#/scraping`	`/scrape`	navin_scraping
Documents	`#/content`	`/studio`	navin_contenant
Marketing	`#/marketing`	`/marketing` `/campaign`	navin_marketing
Montage	`#/montage`	`/montage`	navin_montage
Ads	`#/ads`	`/ads`	navin_ads
SEO	`#/seo`	`/seo`	navin_seo
Tenders	`#/tenders`	`/tenders`	navin_tenders/en
Career	`#/career`	`/career`	navin_career/en
Trading	`#/trading`	`/trading`	navin_trading/en
Leads	`#/leads`	`/leads`	navin_leads
Notes	`#/notes`		navin_notes
Meeting	`#/meeting`	`/meeting`	navin_meeting

Desk loop + heartbeat (Career, Trading, Tenders, Marketing): studio/desk-loop

Settings (source of truth)

Open Ctrl+G in navin-cli (or Settings in the desktop). Sections match navin/tui/settings.py / the desktop Settings view.

Providers

API keys, apiBase, OAuth. Local examples: Ollama, LM Studio, vLLM, any OpenAI-compatible server.

OAuth helpers: navin provider login (openai_codex, github_copilot, xai_oauth).

There is no navin config set api-key. Use this form, an env var, or OAuth.

providers · provider-cookbook · cli/settings

Models

Model configurations, the active default, and task routing (plan / dev / review / security / deep / docs, …). Switch in chat with /model <preset>.

configuration · model fallback

Tools and MCP

MCP servers and presets that extend the agent. Ctrl+U is the same hub.

Configure MCP · cli/tools

Skills

Enable or disable skills. Install from git, npm or a folder. Ctrl+K.

Marketplace · skills evolution

Image / Video / Voice

Generation defaults (provider, model, aspect, size, duration). Voice: STT, realtime, TTS, music.

image-generation · voice

Web

Search provider (DuckDuckGo, Brave, Exa, Tavily), API key, max results, Jina reader for page fetch.

Configure web search

System

Time zone, assistant name, machine resource caps (RAM share, agents per core).

Security

Command confirmation (risky only / always), ask for permissions, remember decisions, security profile (auto / autonomous / assisted / strict), shell on/off, restrict to workspace, builtin deny rules, extra deny/allow regexes, localhost service access.

Ctrl+E opens the security hub.

Guardrails

Per-project board autonomy: consent, auto-branch, PR on done, GitHub issue sync, autopilot loop, kill switches.

Ctrl+Y. board-autonomy

Git

Global auto-branch / PR switches and forge tokens (GitHub, GitLab, Forgejo).

Browser

Playwright: visible window vs headless, live view in the editor.

Rules

Project rules in .navin/rules/*.md.

About

Version, config path, workspace, link to navin.live/docs, raw schema-validated config.json.

Channels and always-on

Connect Telegram, Discord, Slack, Email, Mattermost, Teams, WhatsApp and more.

Chat apps
guides index
Deploy gateway

Safety model

More autonomy does not grant more permissions by itself.

Control	Where
Approvals	Settings → Security
Sandbox	`navin-sandbox` (`make native`)
Checkpoints	agent loop / git
Guardrails	Settings → Guardrails, Ctrl+Y
Kill switches	Guardrails + Git
Rollback	AGI world / policy / skill promote

Environment and ports

navin ports
navin doctor
navin status

Docs index

Everything else is listed in docs/README.md.