Live neural network — all 15 agents, real-time connections, memory, vault
Active model: loading... | checking...
Devin cloud: adaptive · opus · sonnet · swe · gpt · codex · gemini · deepseek · kimi
Active mode: Normal · Cycle with Shift+Tab
/normal
Normal
Auto-approves read-only tools. Asks for write/execute. Default mode.
/accept-edits
Accept Edits
Auto-approves file edits. Still prompts for shell commands.
/plan
Plan
Read-only planning. Explore and ask questions before changes.
/bypass
Bypass
Auto-approves ALL tool calls including writes and shell.
/autonomous
Autonomous
Full autonomous mode. Agent runs independently.
Model Context Protocol servers — extend agent capabilities
Reusable expert knowledge — invoke with /skill-name or agent auto-invokes when relevant
Live terminal — route commands to Intel server :5500
Lexi Terminal — THE SPARK · Agent-005 · Intel :5500
// GOAT Force Terminal — Lexi THE SPARK
// Type a command and hit Run or Enter
lexi@goat-force ~ $
| Shortcut | Action |
| Enter | Submit message |
| Shift+Enter / Alt+Enter | Insert newline (multi-line input) |
| Shift+Tab | Cycle modes: Normal → Accept Edits → Plan → Bypass → Autonomous |
| Ctrl+C | Cancel input / cancel running agent |
| Ctrl+D | Exit (when input is empty) |
| Esc | Cancel running agent |
| Ctrl+G | Open external editor for composing message |
| Ctrl+R | Fuzzy search over previous prompts |
| Ctrl+O | Full-screen viewer for thinking trace |
| Ctrl+L | Redraw / refresh screen |
| Ctrl+V | Paste from clipboard (supports images) |
| ! | Enter bash mode (when input is empty) |
| @ | File/directory autocomplete — add context |
| Opt+T | Cycle thinking level for current model |
Slash commands: /model /mode /plan /ask /bypass /normal /accept-edits /clear /fork /help /bug /update /theme /shortcuts /export /exit
Live global rules — injected into every agent session
Loading rules...
Recent Devin sessions — resume any session
latestLexi — THE SPARK — GOAT Force build
This session · GOAT-Royalty-App · Adaptive
–cContinue most recent session
devin --continue · resumes last session in current directory
⚡ Performance
OLLAMA_NO_CLOUD — keep all inference local
Offline mode — no external downloads or telemetry
Auto-launch all agents on startup
Streaming responses — real-time token output
Cache responses — reuse identical prompts
Auto-continue — resume interrupted responses
🔐 Security
Safe Mode — block unsafe code execution
🎨 Interface
Show line numbers in code blocks
Syntax highlighting in code blocks
Show token counter in toolbar
Performance metrics overlay
14px
🔌 Integrations
Web Search — allow Lexi to search the web
💾 Session Data