Skip to main content

Agents Configuration

Configure how your AI agents behave, what tools they can use, and how they interact with your project.

Agent Panel​

Open the agent panel from the sidebar in any project to configure:

Tools​

Control which tools your agent can use:

  • Enabled tools — Toggle individual tools on or off
  • Tool categories — Enable or disable entire categories (file, terminal, git, deploy, etc.)
  • Auto-approve — Set which tool categories run without confirmation

Memory​

Agent memory determines what context persists across conversations:

Memory TypeDescription
Working memoryShort-term context within a single conversation
Episodic memoryKey facts and decisions remembered across sessions
Semantic memoryEmbeddings-based knowledge from project files

Configure memory settings:

  • Memory retention — How long episodic memories are kept
  • Memory search — Enable semantic search across past conversations
  • Memory panel — View and manage stored memories from the agent drawer

LLM Settings​

SettingDescription
ProviderWhich AI provider to use (OpenAI, Anthropic, etc.)
ModelSpecific model within the provider
Speed tierDefault speed tier (Instant, Fast, Deep, Reasoning)
TemperatureCreativity vs. determinism (0.0 - 1.0)
Max tokensMaximum response length

System Prompt​

Customize the agent's system prompt to control its personality and behavior:

  • Default prompts are provided for common use cases
  • Custom prompts override the default
  • Project-specific context from .chyper/GUIDELINES.md and .chyper/CONTEXT.md is automatically included

Scheduling​

Set up agents to run on a schedule:

  1. Open the agent panel
  2. Go to the Schedule tab
  3. Set a cron expression or use presets (hourly, daily, weekly)
  4. Define the task the agent should perform
  5. Enable the schedule

Schedule Use Cases​

  • Daily code review — Run a security scan every morning
  • Weekly reports — Generate project status reports
  • Continuous monitoring — Check for dependency updates

Execution Plans​

View and manage the agent's execution plans:

  • Plan history — See previous plans and their outcomes
  • Plan details — Inspect individual steps, tool calls, and results
  • DAG Inspector — Visualize the task dependency graph