navvi 3.4.1

Give your AI agent a real browser identity — persistent personas, anti-detect browser, credential vault

Navvi

Give your AI agent a real browser identity.
Persistent browser personas that remember logins, manage credentials, and never get detected.

Quick Start · Use Cases · How It Works · MCP Tools

---

The Problem

Every time your AI agent needs to use the web, it starts from scratch. No cookies, no saved passwords, no history. It has to log in again and again — and half the time the automation gets detected and blocked.

• Agent fills a login form → site detects Selenium/Playwright → blocked
• Agent stores a password in a variable → session ends → password gone
• Agent tries to reuse a browser → cookies wiped → logged out again
• You paste credentials into the chat → now they're in your conversation history

Your agent has no identity. Every session is a stranger.

The Solution

Navvi gives your agent a persistent browser with its own identity. A Camoufox (anti-detect Firefox) that remembers where it's been, stays logged in, and manages its own credentials — without ever exposing passwords to the AI.

• Persistent sessions — cookies, logins, and history survive restarts
• Credential vault — passwords stored in gopass, auto-filled into forms without the AI ever seeing them
• Undetectable input — OS-level mouse and keyboard events (isTrusted: true), navigator.webdriver = false
• Live view — VNC link for when a human needs to step in (CAPTCHAs, OAuth, 2FA)
• Persona management — create multiple browser identities with accounts, credentials, and action history
• MCP-native — Resources, Prompts, and progressive disclosure built into the protocol

Quick Start

1. Build the Docker image

git clone https://github.com/Fellowship-dev/navvi.git
cd navvi
docker build -t navvi:camoufox -f container/Dockerfile container/

2. Add to Claude Code

claude mcp add navvi -- uvx navvi@latest

Or add to your project's .mcp.json:

{
  "mcpServers": {
    "navvi": {
      "command": "uvx",
      "args": ["navvi@latest"]
    }
  }
}

3. Use

Just tell your agent what to do:

"Search DuckDuckGo for 'navvi browser' and list the top results"
"Log into Tutanota with stored credentials"
"Go to github.com/Fellowship-dev/navvi and screenshot the README"

Navvi's journey tools (navvi_browse, navvi_login) handle navigation, element finding, clicking, typing, and screenshots internally. No manual step-by-step needed.

For fine-grained control: atomic tools

Atomic tools are hidden by default. Unlock them when you need precise control:

navvi_atomic(enable=true)                          → unlock low-level tools
navvi_open url=https://example.com                 → navigate
navvi_find selector="input[type=email]"            → locate element → (x, y)
navvi_fill x=512 y=498 value="[e-mail]"     → type into it
navvi_screenshot                                   → see what happened

4. Optional: Install Tatl companion

Tatl is a companion skill that gives Claude Code a dedicated browsing subagent — isolates browser work from your main conversation.

curl -fsSL https://raw.githubusercontent.com/Fellowship-dev/tatl/main/install.sh | bash

Use Cases

Persistent logins. Log into a service once — your agent stays logged in across sessions. No more re-entering credentials, no more expired sessions.

Secure credential management. Passwords live in gopass inside the container. The autofill action types them directly into the browser — the AI never sees the raw password.

Visual evidence for PRs. Screenshot your staging app before and after a code change. Record a user flow as a GIF. Attach it to the pull request.

Form automation on protected sites. Fill complex forms with dropdowns, date pickers, and multi-step wizards. OS-level input passes bot detection that blocks Selenium and Playwright.

Human handoff for hard CAPTCHAs. When the agent hits a CAPTCHA it can't solve, it sends you a VNC link. You solve it in your browser, the agent continues.

How It Works

Your AI agent (Claude Code, etc.)
    |
    | MCP protocol (stdio)
    v
  navvi (FastMCP, Python — via uvx)
    |
    | HTTP → localhost:8024
    v
+--------------------------------------+
|  Docker container                    |
|                                      |
|  Camoufox ←  Marionette (navigate)  |
|     |                                |
|  Xvfb     ←  xdotool (click, type)  |
|     |                                |
|  x11vnc   →  noVNC (live view)       |
|                                      |
|  gopass (credential vault)           |
|  navvi-server (REST API)             |
+--------------------------------------+
    |
    v
  Docker volume (persistent profile)
  ~/.navvi/navvi.db (persona state)

Anti-detection uses Camoufox — a patched Firefox with fingerprint masking at the C++ level. navigator.webdriver returns false.

All input uses xdotool — OS-level events that websites cannot distinguish from a real person.

Credentials are stored in gopass inside the container. The autofill action reads gopass and types directly into the browser — the password never travels through the AI.

Personas persist in Docker named volumes (browser profiles) and SQLite (config, accounts, action logs).

MCP Tools

By default, Navvi shows 11 high-level tools. Atomic tools unlock on demand.

Journey tools (default)

Tool	What it does
navvi_browse	Primary tool — give it an instruction + URL, it handles everything
navvi_login	Log into a service using stored gopass credentials

Lifecycle

Tool	What it does
navvi_start	Start a browser container for a persona
navvi_stop	Stop container (profile preserved)
navvi_status	Show running containers and health

Observation

Tool	What it does
navvi_screenshot	Capture the screen
navvi_vnc	Get live VNC URL for human handoff

Persona management

Tool	What it does
navvi_persona	Create, update, list, delete browser personas
navvi_account	Track accounts per persona (service, email, gopass ref)

Progressive disclosure

Tool	What it does
navvi_atomic	Unlock/hide 12 low-level tools (click, find, fill, etc.)

Atomic tools (hidden by default)

Tool	What it does
navvi_open	Navigate to a URL
navvi_find	Find element by CSS selector → screen (x, y)
navvi_click	Click at coordinates
navvi_fill	Click + type text
navvi_press	Press a key
navvi_scroll	Scroll the page
navvi_drag	Drag between two points
navvi_mousedown/up/move	Low-level mouse control
navvi_url	Get current page URL
navvi_creds	List, get, or autofill gopass credentials
navvi_list	List available Codespaces (remote mode)

Recording tools (hidden by default)

Tool	What it does
navvi_record_start	Start recording screenshots
navvi_record_stop	Assemble MP4
navvi_record_gif	Convert to GIF

MCP Resources

Read persona state without tool calls:

URI	What it returns
personas://list	All personas with account counts
persona://{name}/state	Config, accounts, recent actions
persona://{name}/accounts	Account details
audit://{name}/log	Last 20 actions

MCP Prompts

Structured workflows available as prompt templates:

Prompt	What it does
signup_flow	Step-by-step account creation on a service
login_flow	Log in using stored credentials
qa_walk	Walk a page for QA — screenshot, find issues, report

Personas

Each persona is a separate browser identity with its own cookies, credentials, and history. Managed via MCP tools:

navvi_persona(action="create", name="mybot", description="GitHub admin", stealth="high")
navvi_persona(action="list")
navvi_account(action="add", persona="mybot", service="github.com", email="[e-mail]")

Persona config and state live in ~/.navvi/navvi.db. Browser profiles persist in Docker named volumes (navvi-profile-<name>).

Requirements

• Docker — the browser runs in a container
• uv — curl -LsSf https://astral.sh/uv/install.sh | sh (or brew install uv)
• ffmpeg (optional) — only needed for video recording
• ANTHROPIC_API_KEY (optional) — enables Haiku vision for navvi_browse ($0.002/step). Without it, falls back to heuristics.

License

MIT