Eduardo Machado
Eduardo Machado is an AI Engineer and Software Engineer who builds applied AI systems and open-source developer tools. His current flagship project, Tucano, is a free, open-source desktop/server team of AI agents that collaborates in channels, uses a user's existing AI subscriptions or company gateway, and requires approval before consequential actions.
At a Glance
- Individual developers and technical users
- Software engineering and product teams
- Organizations with internal tools and AI gateways
- Teams seeking local-first, governed agent automation
AI Tools by Eduardo Machado
(1)Tucano
Open Source Multi Agent Desktop App
Discussions
No discussions yet
Be the first to start a discussion about Eduardo Machado
Latest News
Tucano 0.2.8 released with Cursor CLI support, image viewing improvements, customizable agent colors/pictures and drawn avatars, and per-agent routine limits.
Tucano 0.2.7 released with reserve models, multiple Claude Code/Codex accounts, Intel Mac support, and a notification-sounds switch.
Tucano 0.2.5 added controlled use of the user's own desktop applications on macOS, Windows, and Linux through agent-desktop.
Tucano 0.2.4 made Codex model discovery account-driven and improved Windows harness detection and installation.
Products & Services
Free, Apache-2.0 open-source software for running a team of AI agents on a user's Mac, Linux PC, Windows PC, or experimentally on a server. Agents converse in channels, delegate and review work, use MCP-connected tools, run in isolated computers and browsers, follow routines, and pause for approval before consequential operations.
A public repository Eduardo describes as a project created to simplify Linux terminal configuration.
A public Claude Code plugin marketplace project shipping an SDD pipeline (interview-to-spec, diagnose, spec-approach, TDD, and verify) with hooks that turn skill rules into mechanical guarantees.
Market Position
Tucano positions itself as a local-first orchestration and governance layer for multiple existing coding/AI-agent CLIs rather than as a model provider. Its differentiators are team-chat-style multi-agent coordination, per-agent MCP access controls, approval-gated real-world actions, isolated computers/browsers, routines, receipts, and use of subscriptions the customer already has; the practical alternatives are running Claude Code, Codex, Cursor, OpenCode, Pi, or Gemini CLI separately in terminals, or using other multi-agent/developer-agent workspaces.
Founding Story
Eduardo created Tucano to address the mess of running several coding agents across terminals, copying work between sessions, and worrying that an agent might do something it should not. The initial vision was a team-chat-like workspace for multiple agents, with per-agent tool permissions, visible steps, receipts, limits, and human approval for actions that touch the real world.
Business Model
Revenue Model
Free and open-source Apache-2.0 software. Tucano runs agents through the user's existing subscriptions, configured providers, or company AI gateway; its README explicitly says there is no per-token bill from Tucano and no new API key to paste.
Pricing Tiers
Downloadable desktop/server software; users supply their own supported AI-harness subscription, provider login, or company gateway.
Target Markets
- Individual developers and technical users
- Software engineering and product teams
- Organizations with internal tools and AI gateways
- Teams seeking local-first, governed agent automation
- Software teams coordinating several coding, research, review, and automation agents
- Running coding agents against repositories with controlled GitHub/GitLab/Linear/Jira workflows
- Scheduled engineering and operations routines, such as reports, reviews, and monitoring-triggered work
- Working with company-internal tools through MCP and an enterprise AI gateway
- Safe browser and desktop-app automation where publishing, paying, sending, scheduling, or deleting requires human approval
- Personal developers using existing Claude, ChatGPT/Codex, Gemini, Cursor, OpenCode, or Pi subscriptions