Spettro

What

Spettro is Spettro is a terminal-first multi-agent coding assistant written in Go. It targets terminal users, developers, self-hosters, and coding agent users, offering configurable agents, goal mode, agent skills, approval-aware execution, persistent sessions, and OpenAI-compatible model routing.

Key capabilities

• Configurable agents and modes: Cycle the active manifest agent/mode with /mode and /next; the agent system starts with default_agent from the manifest (default: plan).
• Approval flow: When permission is ask-first, a plan is generated first, then /approve executes the pending plan through the coding agent.
• Session compaction: /compact compacts the transcript in two stages — oversized tool results are first replaced with [offloaded: ...] stubs the agent can re-read via tool-output (lossless), then the remainder is summarized.
• MCP management: /mcp <list|read|auth> manages MCP resources and auth.
• Memory editing: /memory edit [user|project] opens a memory file in $EDITOR (default vi).
• Background jobs: /jobs [list] lists background shell jobs started by the agent.
• Subscription management: /login and /logout manage your Spettro Subscription.
• Codebase initialization: /init analyzes the codebase and creates/updates SPETTRO.
• Sandbox read directories: --sandbox-allow-read-dir <dir> adds an extra readable root, useful for tool-chain caches outside the workspace (repeatable).

Platforms

and licensing Spettro supports Linux, macOS, and Windows. It uses a freemium pricing model with a free option. The project is open-source, exposes a public API, and supports self-hosting.