# Booley **The open-source Agentic Development Environment (ADE) for RTL design and verification** [![Tests](https://github.com/boldaxolotl/Booley/actions/workflows/test.yml/badge.svg)](https://github.com/boldaxolotl/Booley/actions/workflows/test.yml) [![PyPI](https://img.shields.io/pypi/v/booley-rtl)](https://pypi.org/project/booley-rtl/) [![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0) [![Python 3.11+](https://img.shields.io/badge/python-3.11+-blue.svg)](https://www.python.org/downloads/) Booley combines the speed of agentic workflows with strict, deterministic EDA checks and thorough reviews by engineers and LLMs. Agents get every tool hardware design needs, safely sandboxed. You get a control panel to orchestrate them and metrics you can trust. **Website:** [booley.dev](https://booley.dev/) for the full feature tour, videos, and blog. Booley in VS Code with two Developer Agent sessions, the Booley
  Dashboard, and waveform inspection ## What it does The [Features page](https://booley.dev/docs/features/) is the complete feature reference. ### Agents equipped for hardware work A general-purpose coding agent knows little about EDA tools, so Booley gives it one short command per job: simulate, synthesize, lint, build for FPGA. The same commands work whichever EDA tool runs underneath, such as Verilator or Icarus Verilog for simulation. It also gets tools for checking its own work. Waveform queries let it ask questions about real simulation traces ("when did `data_o` equal `0xDEADBEEF`?") instead of guessing from the RTL. Mutation testing plants small bugs in the RTL and counts how many the testbench catches, which shows whether the tests would notice a real bug. For a second opinion, the agent can call an independent reviewer that didn't write the code. Reviewers follow guides written by hardware engineers, covering common RTL bugs, style rules and typical optimization opportunities. Each review looks for one kind of problem, so nothing gets skimmed. The agent presses buttons on a Booley panel: Run and Check buttons
  drive the EDA tools, Review buttons go only to a separate Reviewer agent ### Run many agents, stay in control Agents work best when they don't stop to ask permission, and Booley makes that safe. Each agent runs in a container that sees only your project, with internet access limited to the LLM provider. You can leave it running unattended; it can't touch the rest of your machine. The same container gives everyone on the team identical tool versions. Run several agents at once. Booley creates and manages a separate git worktree and branch for each one, so they work independently and never overwrite each other. A built-in scheduler learns how much memory each simulation or synthesis run needs and starts as many together as your machine can hold, so small runs don't wait behind big ones and nothing runs out of memory. Everything sits in one VS Code window: agent chats, RTL, terminals, waveforms, and a dashboard showing what every agent is doing. Inside the container each agent works in its own worktree and
  branch; only the LLM provider is reachable, and you review and merge finished
  work into main ### Results you can trust Normally the result of a task reaches you second-hand: a project-specific script runs the EDA tool, then the agent reads the log and tells you what it found. Each hand-off can go wrong. Booley comes with built-in, tested recipes for each supported EDA tool, built on the open-source FuseSoC. They launch the run and read the numbers out of the tool's reports, the same way on every project. Nobody writes a script for your repo, and no agent interprets a log. Set limits for your design, such as area, timing, coverage or mutation score. An agent works far better against a bound it can check than against "make it smaller". Booley measures every finished commit against all of them, so a change that cuts area but halves Fmax fails instead of slipping through, and the recorded numbers show how the design moved over time. Without Booley, results pass through glue scripts and an agent
  reading logs, and either can go wrong; with Booley, Booley runs the tool and
  extracts the numbers, and the commit carries them marked Measured ## Quick Start Three ways in, ordered by how much you want to invest: 1. **[Watch](#watch).** See an engineer drive Booley on a demo project, start to finish. Zero setup. 2. **[Demo](#demo).** Open the demo free in your browser, or clone it, then type your first request to the agent. 3. **[Your own project](#your-own-project).** Full integration on your own RTL. ### Watch Two videos on the [Videos page](https://booley.dev/videos/) show an engineer driving Booley on a demo project end to end, so you can see the workflow before touching anything. ### Demo Try the demo in browser VS Code with one click. It is free, needs no install, and works on macOS too: [![Open in GitHub Codespaces](https://github.com/codespaces/badge.svg)](https://codespaces.new/boldaxolotl/booley-prj-picorv32?quickstart=1) The Codespace fits in GitHub's free monthly Codespaces allowance, and the agent runs GPT-6 Luna on us after a quick GitHub sign-in. Connect your own Claude or ChatGPT account if you want a smarter model. The first launch takes 3-4 minutes while the container image builds. [How the free demo works](https://booley.dev/demo/). Or **[follow the demo repository's README](https://github.com/boldaxolotl/booley-prj-picorv32#readme)** to run the demo locally, after you [install Booley](#installation). ### Your own project Follow the [Setup guide](https://booley.dev/docs/setup/) to integrate Booley with your own RTL project, after you [install Booley](#installation). ## Installation Booley runs on Windows and Linux (macOS is not supported) for SystemVerilog or Verilog designs. It needs Python 3.11+, Git, Docker, VS Code, and a Claude Code or Codex CLI. With [pipx installed](https://pipx.pypa.io/latest/how-to/install-pipx.html): ```bash pipx install booley-rtl booley bootstrap ``` The [Install guide](https://booley.dev/docs/install/) has the full requirements, upgrade steps, and what `booley bootstrap` does. **Next:** [try the demo](#demo) or [set up your own project](#your-own-project). ## Supported EDA Tools Current integrations: - **Simulate / elaborate** — Verilator, Icarus Verilog; cocotb testbenches supported - **Lint** — Verilator, Verible - **ASIC synthesis** (PPA estimate, not tape-out) — logical Yosys or physical Yosys + OpenROAD - **Waveform debug** — `bwave` (+ VaporView GUI in VS Code) - **FPGA implementation** — AMD Vivado - **Coming soon** — Synopsys VCS, Cadence Xcelium For exact versions, provisioning, trace support, and platform constraints, see **[Supported EDA tools](https://booley.dev/docs/supported-eda-tools/)**. Support for additional commercial EDA tools is coming soon; see the [roadmap](https://github.com/boldaxolotl/Booley/blob/main/docs/internals/ROADMAP.md#commercial-eda-tools). ## Documentation - [Features](https://booley.dev/docs/features/) - [Architecture](https://booley.dev/docs/architecture/) - [Setup](https://booley.dev/docs/setup/) - [Usage](https://booley.dev/docs/usage/) - [Metric-Driven Development](https://booley.dev/docs/metric-driven-development/) - [Flow reference](https://booley.dev/docs/flow-reference/) - [Supported EDA tools](https://booley.dev/docs/supported-eda-tools/) - [Limitations](https://booley.dev/#limitations) - [How Booley is built](https://booley.dev/#how-it-is-built) ## Contributing Booley is still early, so the most useful contribution is trying it and reporting what works, what doesn't, and what you want next. Tell **`/booley-feedback`** in your Developer Agent chat; it gathers and redacts any needed evidence. Opinions need no reproduction, and nothing leaves your machine until you approve the exact text ([feedback guide](https://booley.dev/docs/usage/#when-booley-itself-misbehaves), [configuration](https://booley.dev/docs/config/#feedback-feedback)). Code and documentation contributions are welcome; see [CONTRIBUTING.md](https://github.com/boldaxolotl/Booley/blob/main/docs/internals/CONTRIBUTING.md). Please keep feedback technical and specific; broader debates about AI's effects on society or employment are outside the project's scope. For suspected vulnerabilities, follow the private reporting process in [SECURITY.md](https://github.com/boldaxolotl/Booley/blob/main/SECURITY.md) instead of opening a public issue. ## Acknowledgments Booley stands on a lot of other people's work. Thank you to: - **The authors of [Edalize](https://github.com/olofk/edalize) and [FuseSoC](https://github.com/olofk/fusesoc)**, and especially their lead maintainer, Olof Kindgren, for the framework that makes Booley's whole idea of a simple, unified CLI-over-EDA interface possible. - **The author of [vcdvcd](https://github.com/cirosantilli/vcdvcd), Ciro Santilli**, for the VCD-parsing work that seeded the `bwave` idea. - **The author of [wavepeek](https://github.com/kleverhq/wavepeek)**, another neat waveform-to-CLI EDA tool, for the clean top-level CLI interface that inspired `bwave`'s top-level CLI (the internals started well before wavepeek and are quite different). - **The author of [VaporView](https://github.com/Lramseyer/vaporview), Lloyd Ramseyer**, for the excellent VS Code waveform viewer that `bwave gui` drives for scoped waveform inspection right in VS Code. - **The authors of [Yosys](https://github.com/YosysHQ/yosys), [Verilator](https://github.com/verilator/verilator), [Icarus Verilog](https://github.com/steveicarus/iverilog), [Verible](https://github.com/chipsalliance/verible), and [sv2v](https://github.com/zachjs/sv2v)**, for the excellent open-source EDA tools that make Booley possible at all. - **[Matt Pocock](https://www.aihero.dev/)**, for his great agentic software engineering techniques, which shaped how Booley's agents are built and driven. ## License Apache 2.0. See [LICENSE](https://github.com/boldaxolotl/Booley/blob/main/LICENSE) for details.