Documentation

Tenant OwnerReviewerOperator

Product knowledge base, playbooks, and embedded guidance for every persona.

Agents Overview

Agents are CoffeeBreak’s automated workers.

Each agent knows how to perform a specific kind of task, such as:

  • Implementing a feature
  • Updating or refining a pull request
  • Breaking down an epic into stories and tasks
  • Reviewing code and suggesting improvements
  • Applying changes requested by a human reviewer

You tell CoffeeBreak what you want done.
Task routing and agents handle the “how.”

🧩 How Agents Fit Into CoffeeBreak

When you create a task, CoffeeBreak:

  1. Interprets what you’re asking for
  2. Chooses the right agent (or agents) to handle it
  3. Provides the agent with context (code, docs, history)
  4. Runs the agent to perform the work
  5. Sends the result into review, refinement, or completion

You don’t have to pick an agent manually for most workflows.
Task routing does that for you based on task type and workspace rules.

🧠 Types of Agents

CoffeeBreak typically includes several categories of agents:

1. Implementation Agents

These agents generate or modify code.

Examples:

  • Implementing a new feature from a user story
  • Updating an existing module
  • Adding logging or instrumentation
  • Applying a refactor across multiple files

They focus on creating or changing code in your repositories.

2. Refinement Agents

These agents update an existing change based on feedback.

Examples:

  • Apply requested changes from a human reviewer
  • Address AI Reviewer comments
  • Fix style or naming issues
  • Add or adjust tests

They’re most often used after a pull request already exists.

3. Review & Analysis Agents

These agents help with quality and understanding.

Examples:

  • Summarize a pull request
  • Highlight risky changes
  • Point out missing tests
  • Explain what changed and why

The AI Reviewer is an example of a review-focused agent.

4. Planning & Decomposition Agents

These agents help turn big ideas into smaller pieces of work.

Examples:

  • Break down an epic into features, stories, and tasks
  • Suggest follow-up work items
  • Propose a sequence of steps to implement a larger change

They’re useful early in the lifecycle when you’re still shaping the work.

🧾 Built-In vs Custom Agents

CoffeeBreak includes a set of built-in agents that cover common workflows, such as:

  • Implement/Update feature
  • PR refinement
  • AI code review
  • Epic decomposition

You can also define custom agents tailored to your organization.

Custom agents can:

  • Use your own prompts and instructions
  • Target specific repositories or directories
  • Follow your internal conventions and standards
  • Integrate with internal tools or APIs (depending on your setup)

See:

⚙ Agent Configuration

Workspace admins can control:

  • Which agents are available
  • Which agents are enabled by default
  • Which tasks route to which agents
  • Whether certain agents are restricted to specific repos or branches
  • Whether agents can run automatically or only when triggered

Most teams start with defaults and gradually tune agent behavior over time.

🔍 How Agents Use Context

Agents can use:

  • Your source code
  • Documentation in the Shared Context Store
  • Prior task history
  • Review feedback
  • Workspace configuration

This allows agents to generate work that matches your style and patterns, not just generic examples.

For more details:

🧪 Safeguards and Review

Agents never bypass your existing controls.

CoffeeBreak always respects:

  • Branch protection rules
  • Required reviewers
  • Approval policies
  • CI and test pipelines
  • Workspace confidence and retry rules

Agents can automate the work, but humans remain in control of what gets merged and deployed.