Skip to main content

Terminal UI for managing multiple Claude sessions with tmux

Project description

Cerberus

An orchestrated coding interface where you design, and agents implement.

Cerberus is a multi-agent development environment that lets you focus on the creative work of software design while coordinating swarms of AI agents to handle implementation. You provide the vision, Cerberus manages the execution.

Philosophy

The best coding experiences understand the tango of deference between human and AI thinking.

Humans excel at:

  • Articulating vision and requirements
  • Making architectural decisions
  • Providing context and judgment

AI agents excel at:

  • Implementing well-specified tasks
  • Exploring codebases
  • Handling repetitive transformations

How It Works

Designer and Executor Sessions

Main Session (Designer): Your primary workspace lives in your source code directory. The designer agent discusses with you, understands your requirements, and orchestrates the work.

Executor Sessions: When you define a task, the designer spawns executor agents in isolated git worktrees. Each executor works independently on their assigned task, reaching out to you only when clarification is needed.

The Orchestrator UI

Cerberus provides a unified interface to manage your agent swarm:

  • Session List: View all active designer and executor sessions
  • Diff View: See changes in real-time as agents work
  • Monitor Tab: Watch automated summaries of agent activities
  • Keyboard Controls:
    • s - Open spec editor for the selected session (designer gets notified on save)
    • t - Open terminal in the selected session's worktree
    • N - Create new session
    • D - Delete session
    • Arrow keys - Navigate sessions

Workflow

  1. Design: Discuss your ideas with the designer agent. Use the spec editor (s) to sketch out plans.

  2. Delegate: The designer spawns executor agents with clear task specifications.

  3. Monitor: Watch executors work through the UI. They'll reach out if they need guidance.

  4. Review: Jump into executor worktrees (t) to test changes, or view diffs before merging.

  5. Integrate: Merge completed work back into your main branch.

Architecture

Cerberus uses git worktrees to isolate agent work:

  • Each executor gets its own branch and working directory
  • Changes are tracked independently
  • No conflicts between concurrent agent work
  • Easy to review, test, and merge completed tasks

Communication between agents uses MCP (Model Context Protocol):

  • Designers spawn executors with detailed instructions
  • Executors can message back with questions or completion status
  • Automated monitoring tracks agent activity

Getting Started

# Launch the Cerberus orchestrator
cerb

# The UI will open with a main designer session
# Press 's' to open the spec editor and start planning
# The designer will spawn executors as needed

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

cerb_code-0.1.8.tar.gz (34.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

cerb_code-0.1.8-py3-none-any.whl (29.9 kB view details)

Uploaded Python 3

File details

Details for the file cerb_code-0.1.8.tar.gz.

File metadata

  • Download URL: cerb_code-0.1.8.tar.gz
  • Upload date:
  • Size: 34.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.0

File hashes

Hashes for cerb_code-0.1.8.tar.gz
Algorithm Hash digest
SHA256 d3f92ace5364562b0712212f64dc4dc27eb70f01b180161a454b5db4e7ec53e6
MD5 0dcf826d96c2e02cec6674c4fb0480da
BLAKE2b-256 dbf8b84305a38935990e0564cc311660f5e63d9ee99a591274a9fd18784f26a1

See more details on using hashes here.

File details

Details for the file cerb_code-0.1.8-py3-none-any.whl.

File metadata

  • Download URL: cerb_code-0.1.8-py3-none-any.whl
  • Upload date:
  • Size: 29.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.0

File hashes

Hashes for cerb_code-0.1.8-py3-none-any.whl
Algorithm Hash digest
SHA256 34137dd2539defcb9787720e3b3392baa42135282cea884fd0426134856e7668
MD5 2630fa7a838f4487fba7ffb1be3925de
BLAKE2b-256 65671bc9b85c5d551318e812b9ebb927cf66158b6524e0d4fc2288f87dcfb120

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page