Skip to main content

Operating Agents

Forge provides three specialized agents, each designed for different stages of development work. They differ in capabilities and access levels, allowing you to choose the right approach for your task.

Agent Comparison​

AgentAccessPurposeBest For
MuseRead-writePlanning & analysisReviewing impact, planning changes, critical systems
ForgeRead-writeImplementationMaking changes, fixing bugs, creating features
SageRead-onlyResearch & investigationUnderstanding codebases, tracing bugs, analyzing architecture

Typical workflow: Use Muse to plan → Switch to Forge to implement

Both agents can use Sage internally to research and understand your codebase when needed.

Muse Agent​

Muse analyzes your codebase and creates detailed implementation plans. It proposes solutions and explains the impact of changes without executing them.

Switch to Muse: /muse

Ideal for:

  • Planning complex refactoring
  • Understanding scope before implementation
  • Working with critical or production code
  • Learning how to implement specific features
  • Changes requiring team review

Example prompts:

  • "How would you redesign this API for better scalability?"
  • "Create a plan to add user authentication"
  • "What's needed to implement pagination?"

Forge Agent​

Forge implements solutions directly. It modifies files, creates code, and executes commands to complete tasks immediately.

Switch to Forge: /forge (active by default)

Ideal for:

  • Quick fixes and routine tasks
  • Refactoring with immediate results
  • Implementing approved plans
  • Tasks where you want hands-off execution
  • Creating new features

Example prompts:

  • "Fix the null pointer exception in UserService"
  • "Create a React component for the user profile"
  • "Add unit tests for the payment processor"

Sage (Internal Research Tool)​

Sage is not a user-facing agent, but it's an internal tool that both Muse and Forge can use automatically to research and understand your codebase. When either agent needs to investigate code, trace functionality, or analyze architecture, it leverages Sage behind the scenes.

You don't need to manually switch to Sage; it works transparently when the agents need deeper codebase insights.

Switching Between Agents​

You can switch between agents at any time during your session:

  • Use /muse to switch to Muse Agent
  • Use /forge to switch to Forge Agent
  • Use /agent to see all available agents and choose from a dropdown

Common patterns:

  • Use Muse before making significant changes to critical systems
  • Switch to Forge when you're ready to implement
  • Both agents will automatically use Sage for research when needed

Best practice: Use version control and commit your work before using Forge for significant changes.