Documentation
Everything Spettro does, documented.
The same docs that ship in the repo, rendered here. From first launch to agent swarms, sandboxing, and remote control.
Start here
Getting Started
Install Spettro, trust your project, connect a model, and run your first plan-approve-code loop in the terminal.
Read guideFirst-Launch Onboarding
What happens the first time you launch Spettro with no provider keys: the guided setup wizard, step by step.
Read guideCommands and Keybindings
Complete reference of Spettro slash commands and keyboard shortcuts, from /connect and /models to /goal and /compact.
Read guideConfiguration and Storage
How Spettro stores config, encrypted API keys, trusted paths, sessions, and models under ~/.spettro and per-project .spettro.
Read guideCore concepts
Architecture Overview
How Spettro is built: a Go application with a Bubble Tea TUI front-end, agent runtime, and internal service packages.
Read guideSession Lifecycle
Persistent sessions in Spettro: how conversations are saved, and how /clear, /resume, and /compact keep long threads alive.
Read guideGoal Mode
Run Spettro autonomously with /goal: set an objective, let agents iterate, detect stalls, and stop when the goal is complete.
Read guideUltra Mode (Agent Swarm)
Ultra mode turns the top-level agent into a swarm orchestrator that breaks tasks down and delegates to parallel sub-agents.
Read guideExtended Thinking
Control reasoning budget per model with /thinking levels from off to max, mapped to each provider's thinking API.
Read guideCheckpointing and Rewind
Spettro snapshots your working tree before every change so you can rewind the project to any earlier checkpoint.
Read guideExtend Spettro
Agent Skills
Install reusable capability packs in SKILL.md format from local folders, git URLs, or owner/repo shorthands.
Read guidePersistent Memory
Save facts and preferences across sessions with /memory, with user-scope and project-scope memory files.
Read guideCustom Slash Commands
Write reusable prompts as TOML or markdown files and run them as your own slash commands, with argument placeholders.
Read guideRuntime Hooks
Run custom bash scripts on PreToolUse, PostToolUse, PermissionRequest, and SessionStart to enforce project guardrails.
Read guideSafety & sandboxing
Tools & capabilities
Web Tools
Built-in web-search, web-fetch, and download tools with SSRF-hardened HTTP and an HTML-to-markdown engine.
Read guideVision and Screenshots
Let the agent see images mid-run with view-image and agent-driven screenshots.
Read guideInteractive PTY Sessions
Run commands that need a real terminal - REPLs, TUIs, watchers - in persistent PTY sessions the agent can drive.
Read guideClipboard, Attachments, and Media
Paste images from the clipboard, attach files to prompts, and copy responses with keyboard shortcuts.
Read guideSymbol-Aware Repo Search
How repo-search uses a lightweight symbol index to find definitions and references across your codebase.
Read guideLanguage Server (LSP) Integration
Spettro uses Language Server Protocol servers to give the agent real diagnostics, definitions, and references.
Read guideConnect & integrate
Remote Control Plane
Expose a local HTTP/SSE control plane with /remote so external clients can drive a Spettro session.
Read guideTelegram Relay
Drive a Spettro session from a Telegram chat: send prompts from your phone and get responses back in the chat.
Read guideAgent Client Protocol (ACP)
Run Spettro as an Agent Client Protocol server and use it from ACP-capable editors like Zed.
Read guideSpettro Subscription
How the Spettro subscription works: device-flow login, hosted models under a single key, and plan quotas.
Read guidePlatform notes
Docs are rendered from the spettro repository - edits land here automatically.