Permissions
Velork asks before Agent runs sensitive tools by default. Configure defaults and overrides under Settings → General → AI Permissions, or run Manage AI Permissions from the command palette.
Quick trust presets
| Preset | Effect |
|---|---|
| Trust for this session | Relax prompts for the current session |
| Allow all shell commands in this workspace | Auto-run shell in this folder |
| Allow all shell commands in all workspaces | Global shell auto-run |
Use presets carefully — they apply to every Agent run until you change them.
Permission categories
Each category supports Allow, Ask, or Block:
- File reads
- File edits
- Shell
- Network
- MCP
- Git
Set Default behavior for shell: Ask before running or Auto-run commands.
Command lists
Fine-tune shell access:
- Always confirm before running — extra prompts for specific commands
- Blocked commands — never run
- Never run list — hard block
Path rules
- Block paths — Agent cannot read or write listed paths
- Allow paths — explicit allowlist when using restrictive defaults
Remembered rules
Velork remembers approvals per workspace and session. Review or clear them from Remembered rules tabs — Clear all or Remove individual entries.
Autonomous agent mode
Autonomous agent mode (Cmd + Shift + A) reduces confirmation prompts in chat. Settings → General → AI Permissions also contains an Autonomous agent block for default autonomous behavior — separate from the chat toggle.
Hard Block policies still apply in autonomous mode.
Per-tool prompts in chat
When Agent requests a tool during a run:
- Run Tool — Cmd + Enter
- Skip Tool — Escape
Shell runs may show Run, Skip, Ask before every command, or a link to permission settings.
MCP permissions
MCP tool policies overlap with Settings → Tools & MCP. See MCP servers.
Next steps
Overview
Built-in tools and MCP
Tasks
Automation and validation gates
Settings
Full Velork Settings reference