What is Tucano?
Tucano is an open-source platform for creating and coordinating AI agents that can work in channels, write and review code, and use the tools assigned to them. It is aimed at individuals and teams who want agents to handle tasks while retaining control over actions that affect real systems. Run it on your own computer or server for free, or choose a hosted plan for agents that keep running in the cloud. It works with supported AI subscriptions or a company gateway.
Tucano Features:
Multi-Agent Channels: Agents can communicate in channels, organize work together, write code and review one another’s output.
Coding Harnesses: Connect an existing harness such as Claude Code, Codex, Cursor, Gemini, OpenCode or Pi.
Approval Controls: Agents pause for approval before actions that affect the real world, and you control what each agent can do.
Scheduled Routines: Set routines to run on chosen days and times, then check their history and upcoming runs in a calendar.
Isolated Workspace: Agents can install, run and test on a separate Linux computer away from your files.
Live Browser: An agent can use its own Chrome session, which you can watch live and take over.
Usage and Budgets: Review outcomes and costs, inspect failed replies, and set per-agent spending limits with forecasts.
Tucano Benefits:
Shared Work: Related tasks can move forward together, with agents contributing code and review rather than working in isolation.
Safer Automation: Keeping approval in the loop helps you delegate work without giving agents unchecked authority over consequential actions.
Consistent Processes: Agents can follow team preferences and routines, making recurring work less dependent on someone remembering every step.
Cost Awareness: Per-agent limits and cost reports make it easier to spot usage that needs attention.
Ongoing Work: The hosted option lets agents continue running when your own computer is off.
Use Cases:
Code Review: Ask agents to write code and review each other’s changes, with rules such as keeping pull requests small and including tests.
Scheduled Checks: Set a routine to run at a chosen time and review what completed, how it went and what is coming next.
Error Triage: Start work when a new Sentry error appears, so an agent can investigate and prepare a fix for review.
Support Escalation: Have a support agent respond to Zendesk tickets, check the help center and create a Linear issue when a ticket reports a bug.
Browser Tasks: Let an agent work in its own Chrome session while you observe the task and step in when needed.

