thegent 0.1.0

Unified agent orchestration CLI for Factory skills, droids, and multi-agent workflows

thegent

CLI and framework for agent orchestration, governance, and lifecycle management.

thegent is a CLI and framework for managing AI agent workflows, droids, and multi-agent swarms. It follows a library-first design and uses Rust extensions for performance-sensitive paths.

---

📋 Table of Contents

• Key Features
• Quick Start
• Installation
• Usage
• Performance at Scale
• Governance & Policy
• Security & Hardening
• Documentation
• Docs Deploy
• Contributing
• License

---

✨ Key Features

• ⚡ Performance: Rust-backed tool detection and PATH resolution (<1ms) with 10-100x speedup over shell baselines.
• 🔒 Agent Governance: Built-in policy enforcement, cost caps, and automated quality gates.
• 🌍 Multi-Provider Routing: Routing across Claude, Gemini, OpenAI, and custom local proxies.
• 🛠️ Unified Work Stream: Single source of truth for task management across multiple agents and projects.
• 📦 MCP Native: Full support for Model Context Protocol (MCP) servers and resources.
• 🔄 Continuous Autonomy: Background execution and session management via thegent plan loop.
• 🔍 Deep Research Protocol: Multi-source investigation workflows (Reddit, Google, GitHub).

---

🚀 Quick Start

1. Install (One Command)

macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/kooshapari/thegent/main/scripts/bootstrap.sh | sh -s -- install

Windows (PowerShell):

irm https://raw.githubusercontent.com/kooshapari/thegent/main/scripts/install.ps1 | iex

2. Configure & Verify

thegent install -t all --scope both --full   # Bootstrap user/system assets + provider setup
thegent doctor   # Verify environment health

Project onboarding commands:

thegent scaffold greenfield ./new-project --profile cli_tool
thegent scaffold brownfield ./existing-project
thegent scaffold ag-dd ./existing-project
thegent scaffold none ./existing-project

3. Run Your First Agent

thegent run "Analyze the current directory structure" free

---

📦 Installation

Prerequisites

• Python 3.12+
• Rust (required for building performance-sensitive extensions)
• Homebrew (recommended for system dependencies)

For Developers (From Source)

git clone https://github.com/kooshapari/thegent
cd thegent
pip install -e .
thegent install -t all --scope both --setup
thegent install-shims
thegent setup --hooks

Worktree Governance (Primary-main Flow)

thegent bootstrap and shell/dotfile management support a worktree-first governance model:

• Keep your primary checkout on main.
• Do branch development in dedicated worktrees.
• Merge/cherry-pick branch worktree commits back into main.

When bootstrap runs inside a git repository, it can write a marker file:

• .thegent-primary-main (policy marker)

Interactive shells get a helper function through managed zsh config:

thg_new_worktree <branch> [start-point] [worktree-path]

This helper refuses to branch from a dirty/non-main primary checkout.

Toolchain Setup

thegent uses explicit setup/install surface controls (for runtime assets, shims, hooks, and profiles).
--system-deps, --verify-mise, and --uninstall-mise-hooks are legacy references and are not part of the current parser surface.

For mise/toolchain setup, use one of:

# Use the project-standard bootstrap/install flow first.
thegent install -t all --scope both --setup

# Then manage mise with your preferred shell/toolchain installer separately.
brew install mise
echo 'eval "$(mise activate zsh)"' >> ~/.zshenv

---

🛠️ Usage

Command	Description
thegent run <prompt>	Execute a task in the foreground with a specific agent/model.
thegent run <prompt> --skill <name>	Execute with selected skill instructions (repeat --skill to stack).
thegent bg <prompt>	Start a background agent session.
thegent ps	List active and historical agent sessions.
thegent skill list	List discovered skills available for --skill selection.
thegent skill select <name>	Validate a skill and print exact --skill usage for run flows.
thegent plan loop	Continuously process work items from the unified work stream.
thegent plan do-next	Find the next actionable items from project plans and specs.
thegent doctor	Verify environment health and fix performance bottlenecks.
thegent sync autopilot	Automatic bi-directional sync: WORK_STREAM.md <-> GitHub Projects <-> Linear.

Harness wrappers (dex, clode, roid, droid) route through thegent-shims. Use --native to bypass wrapper-injected defaults/proxy routing and call the underlying native CLI directly.

Skill UX examples:

thegent skill list
thegent skill select thegent-skills
thegent run agent "run with selected skill" --skill thegent-skills

Unknown skill handling is explicit and non-silent:

thegent skill select missing-skill
# Skill not found: missing-skill

Workstream Autosync (GitHub Projects + Linear)

Enable fully automatic reflections so agents can stay unaware of board plumbing:

THGENT_WORKSTREAM_AUTOSYNC_ENABLED=1
THGENT_WORKSTREAM_AUTOSYNC_INTERVAL_SEC=60

THGENT_GH_PROJECT_SYNC_ENABLED=1
THGENT_GH_PROJECT_OWNER=<org-or-user>
THGENT_GH_PROJECT_NUMBER=<project-number>

THGENT_LINEAR_SYNC_ENABLED=1
THGENT_LINEAR_API_KEY=<linear-api-key>
THGENT_LINEAR_TEAM_ID=<linear-team-id>

Run once:

thegent sync autopilot --once

Run continuously:

thegent sync autopilot --interval 60

Task entrypoints:

task sync:autopilot
task sync:autopilot:once

---

📊 Performance at Scale

Operation	Legacy (Shell)	thegent (Rust)	Improvement
Tool Detection	60ms	1ms	60x
PATH Resolution	20ms	0.5ms	40x
Process Scanning	50ms	0.5ms	100x
Hook Execution	200ms	20ms	10x

---

🛡 Governance & Policy

thegent treats AI agency as a governed resource:

1. Cost Control: Define per-session and per-project token/dollar budgets.
2. Quality Gates: Automatic validation of agent outputs against defined specifications.
3. Policy Enforcement: Centralized governance/ module for enforcing security and ethical constraints.
4. Audit Logs: Full traceability of agent actions, including tool use and thought processes.

---

🔐 Security & Hardening

Security controls for agentic operations:

• Minimal Surface: Core logic isolated in Rust for performance and security.
• Stealth Scrapers: Built-in mechanisms to bypass scraping blocks and protect agent anonymity.
• Path Isolation: Strict control over the execution environment via optimized shims.
• Secret Management: Secure storage for API keys and provider credentials.

---

📚 Documentation

• Public Docsite — VitePress-powered public documentation. Run locally with bun run dev from docs/site/. See docs/site/README.md for full setup instructions.
• Docsets — Audience-based documentation tracks.
  • Developer (Internal)
  • Developer (External)
  • Technical User
  • Agent Operator
• CLIProxyAPI Issue Board — 961 tracked GitHub issues from CLIProxyAPI/Plus with thegent solutions.
• Quick Start Guide — Get up and running in 5 minutes.
• Complete User Guide — Deep dive into features.
• Installation Guide — Advanced setup options. <<<<<<< HEAD
• Provider Setup Guide — cliproxy login, provider/model routing, adapter vs native behavior, and troubleshooting.
• Changelog — Keep-a-Changelog release history with active Unreleased section.
• Changelog Process — How to add, classify, and release changelog entries.
• Changelog Entry Template — Copy/paste template and writing guidance for entries.
• Project Setup Style — Standardized command/process baseline inspired by vercel/ai. <<<<<<< HEAD =======
• Provider Setup Guide — cliproxy login, provider/model routing, adapter vs native behavior, GLM + MiniMax flows, and OpenCode/Zen integration notes.
• Domain Mapping Guide — thegent domain map advisor mode for Porkbun + Cloudflare Tunnel domain exposure.
• Release Supply Chain Controls — SBOM, vulnerability scans, governance attestation, and release provenance artifacts.

codex/governance-wireup ======= codex/provider-plane-wave1

• Architecture Overview — Design layers and internals.
• Research Index — Findings and experiments.

---

🚢 Docs Deploy

Local docs:

bun run docs:dev
bun run docs:build

GitHub Pages:

• Workflow: .github/workflows/docs.yml
• URL convention: https://<owner>.github.io/thegent/

---

🤝 Contributing

We welcome community contributions! Please see our CONTRIBUTING.md for:

• Development environment setup (using uv).
• Test suite execution (task test).
• Coding standards and PR process.

---

📜 License

Distributed under the MIT License. See LICENSE for more information.

---

Built with ❤️ by the community