Gobii

What is Gobii?

Gobii was a cloud platform for AI agents that used a real browser to finish repetitive web tasks. You briefed an agent in plain language, and it researched prospects, watched pages or updated spreadsheets on a schedule.

Gobii, Inc. was founded by Andrew Christianson and launched in June 2025, built on the open-source browser-use framework. In August 2026 it posted a wind-down notice, with 1 September 2026 reported as the final day. The MIT-licensed code stays public on GitHub.

Key Features

• Always-on agents: Gobii workers ran in the cloud and woke on schedules, webhooks, emails, SMS or API calls. Agents kept their state across turns, so a recurring task picked up where the last run ended.
• Browser automation: Agents used a real browser to read pages, click buttons and submit forms. The engine was built on the open-source browser-use framework, letting agents work on sites that offer no API.
• Email and SMS endpoints: Each agent could get its own email address and phone number. Colleagues could message a worker directly, and inbound messages woke it to start work.
• Worker templates: Ready-made roles such as Lead Hunter gave a quick start for common jobs. Templates covered sales, recruiting and compliance monitoring, so a new user could adjust an existing brief instead of writing one from scratch.
• Developer API and MCP: A REST API and a hosted MCP server with 16 tools let engineering teams manage agents from their own code. Webhooks passed results on to other systems.
• Agent collaboration: Agents could message each other and hand results on. Multi-step jobs, such as prospect research followed by CRM updates, could be split between specialised workers that each focused on one part of the task.
• Security controls: The open-source code describes encrypted-at-rest secrets, proxy-governed web access and sandboxed compute. These limits reduced the risk posed by agents logged into real business accounts, which matters for anyone self-hosting it now.
• Human-in-the-loop controls: Capterra listed human-in-the-loop controls among Gobii’s capabilities. Teams could keep people involved in agent work rather than relying only on fully autonomous runs, according to that listing.

Pros

• Plain-language setup: Agents were briefed in everyday language, with no scripting needed. Sales and operations staff could build and adjust workers themselves without waiting for developer help or learning a workflow editor.
• Open-source core: The MIT-licensed gobii-platform code remains public on GitHub. Technical teams can still self-host it after the hosted shutdown, although the Gobii name and logo stay trademarked, so copies need their own branding.
• Wide use cases: Templates covered lead research, candidate sourcing, price monitoring and compliance checks. One platform could therefore serve sales, recruiting and operations teams with separate workers sharing the same account.
• Clear entry price: Pro started at a reported $50 a month with 1,000 tasks, and Capterra noted a seven-day free trial. Extra tasks reportedly cost $0.10 each, which made small pilots cheap.
• Partial migration export: Users could reportedly export agent instructions, memory, conversation history and files before closure. That softened the shutdown for teams moving to self-hosting or another tool.

Cons

• Service closed: The hosted product reportedly stopped on 1 September 2026 after a wind-down notice in August. Former users must self-host the open-source code or move to another platform such as Manus or Zapier Agents.
• Variable credit use: One review found higher settings used more credits per task. Monthly costs were therefore hard to predict for heavy or complex workloads, especially on plans that charged per extra task.
• Incomplete export: Credentials, channel settings and schedules were reportedly left out of the migration export. Download links reportedly lasted only seven days, and rebuilding the missing pieces takes time on any new platform.
• Self-hosting effort: Running the open-source edition needs your own servers, LLM API keys and email or SMS providers. Small teams without engineers may find this impractical and should budget for hosting and model usage.
• Few user reviews: Capterra showed no user reviews for Gobii. Independent evidence of reliability stayed limited during the product’s short life, so most published opinions came from the company or directory summaries.

Who is Using Gobii?

• Sales teams: Lead research and CRM enrichment.
• Recruiters: Candidate sourcing from public profiles.
• Operations staff: Price checks, data entry and reports.
• Compliance analysts: Monitoring policy pages for changes.

Pricing

Gobii no longer sells hosted plans; the last reported rate card is below (checked October 2026).

Older directory listings still show a free hosted tier and a 500-task Pro plan. Refunds were reportedly promised to active subscribers, so check with the company directly.

What Makes Gobii Unique?

Each agent could have its own email address and phone number. Many browser agents waited in a chat window, but Gobii workers could be emailed or texted like a colleague. The MIT licence also limited lock-in, which matters now the hosted service has gone.

Compatibilities and Integrations

• Platforms: Web app; the open-source edition runs on Linux servers.
• Business apps: Directories list Gmail, Slack, HubSpot and Salesforce.
• API and MCP: REST API, webhooks and MCP for custom systems.
• Data export: Instructions, memory, history and files were exportable.

Gobii Tutorials

With the hosted app closed, these steps cover moving to the open-source edition.

1. Download any export you still have.
2. Clone the gobii-platform repository from GitHub.
3. Add your own LLM keys and email or SMS providers.
4. Import agents, then rebuild schedules.
5. Test each agent on small tasks.

How We Rated It

Summary

Gobii suited small teams that wanted browser agents they could brief in plain English. Its main limitation is that the hosted service has reportedly closed. We would point most former users to a maintained alternative. If you self-host, test each agent before giving it live data.