Gobii Platform

The production platform for browser-use agents 🚀

Gobii is the open-source platform for deploying and managing browser-use agents at scale. While browser-use gives AI agents powerful browser automation capabilities, Gobii provides the infrastructure to run them in production: always-on execution, scheduling, email/API triggers, secret management, and team collaboration. Spin it up with Docker Compose, complete a first-run wizard, and you have self-hosted browser-use agents that work 24/7. Prefer managed hosting? Gobii Cloud at gobii.ai delivers the same platform as a service.

chat.mp4

Gobii Agent demo in action

What Makes Gobii Different

• Production infrastructure for browser-use: Turn browser-use agents into always-on services with scheduling, email triggers, API endpoints, and persistent execution.
• Self-hosted or managed: MIT-licensed platform you can run anywhere, or Gobii Cloud for zero-ops hosting with SLAs.
• Built for teams: Share agents, manage secrets, collaborate on workflows, and control access across your organization.

Launch in Minutes

1. Prerequisites: Docker with at least 12 GB RAM allocated to its VM and a few GB of disk.
2. Clone & enter the repo

   git clone https://github.com/gobii-ai/gobii-platform.git
   cd gobii-platform

3. Start Gobii (first run and whenever dependencies change)

   docker compose up --build

4. Visit http://localhost:8000 and follow the first-run wizard:
   • Create the first admin account.
   • Pick the LLM provider (OpenAI, OpenRouter, Anthropic, Fireworks, or custom) for your agents.
   • Drop in the API keys and preferred models. You can route browser-use agent calls to a different model if you'd like.
5. After the redirect, sign in at http://localhost:8000/ and start deploying browser-use agents.

Need scheduling, inbox listeners, or extra telemetry later? Launch the optional beat, email, or obs profiles with docker compose --profile <name> up.

What You Can Build

• Deploy browser-use agents with persistent execution, secret management, and email/web chat interfaces.
• Expose browser-use capabilities via API endpoints for teammates or external services.
• Monitor agent execution, capture structured outputs, and manage files generated during workflows.

Try the API

curl --no-buffer \
  -H "X-Api-Key: $GOBII_API_KEY" \
  -H "Content-Type: application/json" \
  -X POST http://localhost:8000/api/v1/tasks/browser-use/ \
  -d '{
        "prompt": "Visit https://news.ycombinator.com and return the top headline",
        "wait": 60,
        "output_schema": {
          "type": "object",
          "properties": {"headline": {"type": "string"}},
          "required": ["headline"],
          "additionalProperties": false
        }
      }'

Choose Your Path

Self-Host (this repo)	Gobii Cloud (managed)
MIT-licensed core, data stays on your infra.	Zero-ops hosting, governed releases, SLAs.
Customize runtime, networking, branding, and integrations.	Autoscaling agents, managed upgrades, enterprise support.
Community support via GitHub issues & Discord.	Dedicated success and support (contracted).

The Platform for browser-use

browser-use is the leading open-source library for AI-powered browser automation, giving agents human-like web interaction capabilities. Gobii provides the production infrastructure to deploy and scale these agents:

browser-use brings the automation:

• Advanced web navigation, form filling, and data extraction
• Visual understanding of web pages for smarter interactions
• Robust error handling and retry mechanisms

Gobii brings the platform:

• Always-on agent execution and scheduling
• Email triggers, API endpoints, and web interfaces
• Secret management, team collaboration, and access control
• Monitoring, logging, and structured output capture

Want to contribute to the browser automation layer? Check out browser-use on GitHub.

Developing

• Start infrastructure locally with docker compose -f docker-compose.dev.yaml up (Postgres, Redis, MinIO).
• Run Django via uv run uvicorn config.asgi:application --reload --host 0.0.0.0 --port 8000 and the Celery worker on macOS with uv run celery -A config worker -l info --pool=threads --concurrency=4 (threads avoids recent fork restrictions).
• Full local workflow, optional profiles, and testing steps live in DEVELOPMENT.md.

Contribute & Connect

• Share ideas or bugs in GitHub issues.
• Follow existing style (ruff/black) when submitting PRs.
• Join the community on Discord.

License & Trademarks

• Source code ships under the MIT License.
• The Gobii name and logo are trademarks of Gobii, Inc. See NOTICE for guidance.
• Proprietary mode and non-MIT components require a commercial agreement with Gobii, Inc.

---

Built with ❤️ by the Gobii team. The production platform for browser-use agents.