Sypha AI Docs
Features

Automated Action Approvals

Optimize your development speed by granting Sypha permission to execute trusted operations autonomously.

Automated Action Approvals

By configuring autonomous approval settings, you enable Sypha to handle routine operations - such as file reading or command execution - without requiring manual confirmation for every step. This streamlines your workflow but necessitates a balanced approach to security.

[!CAUTION] PROTECTION ADVISORY: Granting autonomous access permits Sypha to interact directly with your environment. Executive actions are especially critical. Ensure you only permit automated execution for operations and command structures you thoroughly trust.

Instant Management

The Auto-Approve Toolbar, located above the chat input area, allows you to manage permissions dynamically:

  • Read: View file contents and directory hierarchies.
  • Write: Edit existing files or generate new ones.
  • Execute: Run authorized terminal instructions.
  • Browser: Manage headless browser sessions.
  • MCP: Coordinate with external Model Context Protocol servers.
  • Workflow: Update progress trackers, adjust roles, or manage subtasks.

Permission Risk Profile

Permission TypeRisk LevelImplementation Advice
Read OperationsLow/MediumGenerally safe for reduced interruption.
File ModificationsHighConsider a brief confirmation delay for diagnostic transparency.
Shell ExecutionHighPrimarily use the Allowlist for standard toolchains (e.g., git, npm).
DeletionCriticalAvoid automation here unless operating in a sandbox.

Advanced Security Controls

High-Autonomy Mode (YOLO)

The master checkbox at the start of the toolbar toggles "YOLO Mode." This provides the AI with maximum independence, which significantly boosts speed but requires the highest level of environmental trust.

AI Integrity Guard

For expert users operating with high autonomy, the AI Integrity Guard provides an additional layer of protection. This feature utilizes a secondary, low-latency model to inspect proposed actions for security risks before they are finalized.

Managing the Allowlist

Navigate to Settings -> Auto-Approve to define specific command patterns that are always permitted (such as ls -R or npm test). You may also establish a Restriction List for operations that must always require human oversight.

On this page