Paseo

Paseo is an LLM coding agent orchestrator and multi-agent workflow manager designed to coordinate multiple AI agents across isolated git worktrees. It provides a unified control interface for managing these agents and their associated environments to execute complex programming tasks.

The system distinguishes itself through a remote agent daemon that enables secure access to local coding agents via encrypted relays. It employs a git worktree environment manager to isolate parallel tasks into dedicated directories and branch-based server URLs, preventing file collisions and network port conflicts between concurrent agents.

The platform covers wide-ranging capabilities including multi-agent orchestration via specialized agent committees, iterative worker-verifier execution loops, and comprehensive git workflow management. It includes tools for visual code review, GitHub API integration, and a command line interface for streaming real-time output and managing agent sessions.

The architecture utilizes a headless daemon and a standardized JSON-RPC protocol to communicate with agent binaries over stdio.

Features

Coding Agent Orchestrators - Coordinates multiple AI coding agents using isolated worktrees and terminal session management.

Multi-Provider Orchestrators - Orchestrates AI coding agents from different providers concurrently within the same workspace while maintaining native configurations.

Git Workspace Isolations - Uses git worktrees to create isolated directories and branches for parallel agent tasks to prevent file collisions.

Agentic Workflow Orchestration - Decomposes complex tasks into parallel work streams by spawning sub-agents and synthesizing their results.

Agent Client Protocols - Uses a standardized JSON-RPC protocol over stdio to communicate with agent binaries and manage their execution.

Worktree Isolation - Isolates parallel agent tasks using dedicated git worktrees to prevent file collisions and state interference.

AI Provider Integrations - Provides a connection system for linking various large language model providers using API keys or self-hosted endpoints.

Agentic Workflow Automation - Decomposes complex tasks into parallel streams and iterative loops using hierarchical agent structures.

Provider Mappings - Maps coding agents to specific API endpoints or self-hosted environments via extensible provider definitions.

Hierarchical Agent Orchestration - Implements a structural hierarchy where manager agents decompose tasks into parallel streams for worker agents.

Multi-Agent Task Orchestrators - Manages complex, multi-step workflows by decomposing programming tasks into parallel streams and iterative loops.

Remote Agent Daemons - Runs a background service that enables secure remote access to local coding agents via encrypted relays.

Remote Agent Management - Controls AI agents running on remote servers or background daemons via encrypted relays.

AI Agent Session Managers - Manages the full lifecycle of AI coding agent sessions, including task prompts, status tracking, and termination.

Worktree Isolation Tools - Isolates parallel agent tasks into dedicated directories and branch-based URLs using git worktrees.

Git Workflows - Manages branching and pull requests using isolated git worktrees to enable parallel agent workflows.

Agent Command Line Interfaces - Provides a terminal interface for managing agent sessions, worktrees, and streaming real-time output.

Encrypted Relaying - Links remote clients to a local daemon using end-to-end encrypted relays for secure cross-device control.

Remote Access - Enables secure connectivity to agents on local machines or remote servers via encrypted relays.

Branch-Based URL Routing - Assigns unique, branch-based URLs to parallel agents to prevent network port conflicts.

Agent Daemon Connections - Enables secure, encrypted connectivity to remote agent daemons via local sockets or relay URLs.

Iterative Execution Loops - Employs an automated cycle of agent code generation followed by command verification to ensure success.

Committee Planning - Forms groups of high-reasoning agents to perform strategic planning and root cause analysis for complex problems.

Agentic Execution Loops - Automates the repetition of agent executions until a specific verification command passes.

Task Hand-off Transfers - Implements a transfer mechanism to hand off tasks between agents via briefings containing context and acceptance criteria.

Model List Management - Provides administrative tools for adding, relabeling, and refining the list of available AI models.

Headless Execution Pipelines - Operates as a background orchestration engine on servers without a graphical user interface.

Generation Steering Instructions - Implements a per-project prompt system to steer the wording and format of generated titles and commit messages.

Advisor Solicitation - Allows the system to spawn separate agents to provide a second opinion or judgment on specific tasks.

Agent Preference Settings - Allows users to specify preferred AI models and reasoning levels for background metadata tasks.

Workspace Workflow Scripting - Executes custom shell scripts during workspace creation and archival for dependency installation.

Background Service Management - Manages long-running processes within workspaces and routes HTTP traffic via a reverse proxy.

CLI Agent Management - Includes command-line utilities for controlling agent sessions, configuration, and managing input streams via native adapters.

Pre-Commit Visual Reviews - Includes a visual diff viewer for inspecting modifications to the codebase before committing changes.

GitHub API Integrations - Provides a centralized interface for performing commits, pushes, and pull request management via GitHub APIs.

GitHub Workflow Integrations - Offers an integrated interface for managing repository tasks, commits, and pull requests within the development environment.

Real-Time Output Streaming - Streams the standard output of running agents in real-time to the terminal for progress tracking.

Agent Task Dependency Resolvers - Provides a blocking execution mechanism to ensure agent task dependencies are completed in the correct order.

Session Automations - Automatically launches pre-defined terminal sessions upon workspace creation for real-time logging.

Terminal Session Lifecycle Management - Manages terminal instances for specific directories by capturing output and sending key inputs.

Version Control Integrations - Surfaces pull requests, checks, and reviews within the management interface to streamline the code review process.

Visual Layouts - Provides a visual arrangement system using split panels and tabbed groups for agents, browsers, and terminals.

Daemon Lifecycle Management - Provides tools to start, stop, and monitor the lifecycle of the agent hosting daemon.

Branch-Based URL Aliases - Assigns unique development server URLs based on git branch names to prevent network port conflicts between agents.

JSON-RPC Implementations - Uses a standardized JSON-RPC protocol over stdio to exchange commands and capabilities with agent binaries.

AI Provider Profiles - Supports creating multiple distinct profiles for the same provider to manage different credentials and model lists.

Headless Daemon Management - Provides management of the orchestration engine running as a non-interactive background service.

AI Agent Activity Monitors - Provides a real-time stream and filter for execution history to track agent progress and tool usage.

getpaseopaseo

Features

Star history