Vicoa

You use Vicoa to run a team of agents on a single project, each in its own isolated git worktree so they do not conflict. Its main strength is as a central command center, letting you start jobs on your desktop and check in from your phone, getting notifications when an agent needs input. Because it runs agents directly on your machine without a sandbox, there is a risk of unintended changes if you are not carefully monitoring its actions.

Pick Vicoa if you want to orchestrate multiple agents and value the ability to monitor and interact with them from any device. Pick a single agent like Aider or Cursor if you prefer a simpler one-at-a-time workflow inside a single terminal or editor.

Vicoa calls itself an 'agentic IDE'. It executes code and git commands directly on the host machine. There is no documented sandbox. A misbehaving or compromised agent could corrupt the local environment. The architecture uses git worktrees to isolate parallel agent tasks, which prevents them from overwriting each other's work.

This design trades safety for direct file system access. It is an effective multi-agent harness if you trust the agents you run. The cross-device control from a mobile app is a unique and well-executed feature for managing long-running tasks.

Vicoa is documented as an orchestrator, not a self-contained coding agent. Its primary function is to manage and interface with other existing agents, such as Claude Code and GitHub Copilot, from a unified interface available on desktop, web, and mobile. The architecture isolates each agent session in its own git worktree, which allows for parallel work on a single repository without direct conflict. This design is notable.

The tool's value is dependent on the capabilities of the third-party agents it runs. Since Vicoa provides the management layer but not the agents themselves, its utility is a function of the user's existing subscriptions. The lack of a sandboxed execution environment means agents run with local user permissions, a design choice with clear security implications.

Vicoa is for the power user running multiple coding agents simultaneously. Its core value is orchestration: parallel git worktrees for each agent and cross-device session steering are genuinely useful workflows. This is a solid product for coordinating complex, multi-agent tasks on a single repository without the chaos of conflicting changes.

The business model, however, is a question mark. It's open-source and bring-your-own-key, which signals a classic 'build audience first' strategy. Without a clear path to revenue or a strong moat beyond its interface, the company is a feature waiting for its acquirer. The most likely exit is to a platform that wants to own the multi-agent developer experience, like GitHub or a cloud provider.

This is an agent orchestrator, not a self-contained agent. The ability to manage multiple agent sessions from a mobile device is a novel interface. However, the core product lacks enterprise fundamentals. There is no mention of SSO, SCIM, audit logs, or a commercial license to supersede the AGPL-3.0, which presents a significant adoption barrier for our use case.

The bring-your-own-key model shifts cost to metered API usage, which requires centralized tracking we would need to build ourselves. Without a supported enterprise version, the security, compliance, and legal overhead is too high. It's a tool for an individual, not a team of sixty.

Vicoa is a control plane for managing other coding agents. Its main idea is running multiple agents in parallel, each on its own git worktree, and letting you steer them from a desktop, web, or mobile UI. This is for people running fleets of agents who are tired of juggling terminal tabs. The architecture relies on their relay service to sync commands between your devices and the machines running the agents.

The AGPL license is a good sign for longevity, and bringing your own keys for the underlying agents keeps costs down. However, it's not a self-contained system; it depends on a central Vicoa relay service to connect your devices. If that service goes down, you lose the cross-device control, which is the whole point. Also, no Docker sandbox means agents run with your user's permissions, which is a risk.