CodeWhale

CodeWhale is an AI coding agent orchestrator and development harness designed to coordinate autonomous agents that read, edit, and verify code. It provides a secure environment for AI agents to perform multi-step software engineering tasks, utilizing a sandboxed execution model to isolate shell commands and protect the host system.

The system distinguishes itself by spawning multiple independent agents in parallel to handle separate investigation or implementation slices simultaneously. It employs a multi-model gateway to route requests across various cloud APIs and local servers, and utilizes a hierarchical instruction system to resolve conflicts between global laws, project invariants, and user requests.

The platform covers a broad range of automation capabilities, including autonomous goal execution, reusable workflow loading, and session persistence to resume long-running tasks. It also supports snapshot-based state rollbacks to restore codebases without altering git history and integrates with external protocols to share tool definitions across different AI environments.

The project includes a command line interface for executing autonomous workflows in a headless mode.

Features

AI Agent Orchestrators - Provides a system to organize and coordinate specialized AI agents to read, edit, and verify code through multi-step tasks.

Concurrent Agent Execution - Implements mechanisms for running multiple independent agents in parallel to handle separate implementation slices simultaneously.

Agent Harnesses - Provides a structured development harness for creating and managing autonomous agents that modify code.

Autonomous Coding Agents - Builds and manages AI agents that autonomously decompose complex coding tasks into actionable steps within a terminal.

AI Execution Sandboxes - Isolates shell commands and tool calls from AI agents using OS-level sandboxing to protect the host system.

Hierarchical Instruction Resolvers - Resolves priority conflicts between global laws, project invariants, and user requests using a hierarchical instruction system.

Multi-step Goal Execution - Pursues defined objectives across multiple turns of reading, editing, and verifying until a complex task is completed.

Parallel AI Workflows - Enables the simultaneous execution of multiple independent AI agents to handle separate implementation slices of a project.

Agent Sandboxes - Ships a secure execution environment with OS-level isolation for AI agents to run shell commands and modify files.

Agentic Task Orchestration - Implements autonomous agent workflows for planning, tool invocation, and iterative execution of complex coding tasks.

Isolated Execution Sandboxes - Provides an isolated execution sandbox for shell commands to protect the host system and control access.

Agent State Persistence - Persists the execution state of autonomous agents to ensure continuity across system restarts or sleep.

Model Context Protocol - Integrates with the Model Context Protocol to share tool definitions and execution capabilities across different AI environments.

AI Agent Tool Integrations - Connects AI models to external protocol servers to share and execute specialized tool definitions.

Model Request Routing - Routes requests across diverse AI backends, including hosted cloud APIs and local runtime servers, for flexible model access.

AI Tooling Protocols - Interfaces with external protocol servers to synchronize tool definitions and capabilities across AI environments.

LLM Gateways - Acts as a centralized interface for aggregating and routing requests to multiple cloud and local LLM providers.

Multi-Model AI Orchestrators - Provides a gateway to route requests across diverse cloud APIs and local AI model providers.

Development State Snapshots - Captures temporary project snapshots to restore the codebase to a previous turn without altering git history.

AI Task Runners - Provides a command line interface for executing autonomous AI workflows and tool calls within a sandboxed environment.

Agent Action Rollbacks - Restores the project state to a prior turn using snapshots to revert changes made by autonomous agents.

Agentic Session Persistence - Saves agent contexts and tool call sequences to disk to enable resuming long-running tasks after restarts.

Agent Execution Modes - Offers a command line interface to execute autonomous AI workflows in a non-interactive headless mode.

HmbownCodeWhale

Features

Star history