Versions:
Agent Console is a minimalist, AI-native desktop application developed by Carlos Castillo for directing coding agents inside a software repository. Built with a terminal-first philosophy, it pairs popular coding-agent command-line interfaces — specifically Claude Code and Codex — with an integrated PTY terminal, a git diff viewer, per-turn working-tree snapshots, and per-tool approvals. This combination allows developers to drive an agent through real development work while retaining visibility and control over every action it takes, making it particularly suited to AI-assisted coding, supervised agent workflows, and repository-level automation tasks. A distinguishing feature of the application is its evidence model: every session is witnessed into a hash-chained evidence ledger called Testigo, which can be exported as signed proof packets that are verifiable in any browser, supporting use cases where auditability, accountability, or cryptographic proof of an agent's activity is required. Within the software catalog, Agent Console fits into the developer tools and AI-assisted coding category, alongside agent orchestration utilities and terminal applications. The current release is version 0.79.0, and the project has published eight versions to date, reflecting an active but still pre-1.0 development history. Because Agent Console operates as a console around existing agent tooling rather than a self-contained agent, it has external dependencies: the Claude Code CLI must be installed for the integration to function, and the application requires Node.js 20 or newer as its runtime environment. Overall, Agent Console provides a focused control surface for practitioners who want the productivity of coding agents combined with granular approval gates, snapshot-based recovery points, and tamper-evident session records, all presented through an intentionally minimal interface that keeps the terminal at the center of the workflow.
Tags: