Available
Fleet
One session, a whole fleet. Subagents that work in parallel, one shared task list, and rules every agent follows. A Claude Code plugin, or a plain MCP server for other agents.
ToolsEnabled
Real tools for AI agents, under your control.
On your computer, for the agent you already run.
The ToolsEnabled family
Everything we make.
Ten projects for working with AI agents.
Available
Desktop
Run and organize your AI agents in one desktop workspace, with chats, tools and permissions under your control.
Available
Engine Bench
Two AI models get the same prompt and the same CAD tools, then race to build a real engine, part by part.
Available
Head Gasket Bench
Part two of the engine benchmark. An AI agent replaces the head gasket on the same engine through 18 MCP tools, and a scorer checks the saved record.
Available
Site Studio
Edit a static website in a private working copy, with a live preview and human-controlled export.
Available
Slides
Build editable PowerPoint decks in a local browser studio. Export PPTX, or PDF and PNG with LibreOffice on Linux.
Available
Sim
A shared warehouse for three or four wheeled robots, coordinated through MCP. Run on Rapier or MuJoCo and verify the replay.
Available
ToolsEngine
A local MCP server for browser automation, workspace files and public web reading. 27 tools on Windows and Linux, no shell, and a fresh browser profile by default.
Available
Data
A local workspace to inspect, check and transform CSV or TSV data, then save tables, charts and reports without changing source files.
Available · 0.3.3 beta
ToolsEnabled Bench
A local, generic benchmark builder. Compose tasks, freeze a protocol and regenerate reports from retained evidence. Separate from Engine Bench and Head Gasket Bench.
One benchmark family
Three benches. One way to score an agent.
Engine Bench, Head Gasket Bench, and ToolsEnabled Bench are one family. Each keeps its own name, page, tasks, and scoring.
Start with Fleet
One session.
A whole fleet.
In Claude Code, Fleet is a plugin: add it from GitHub with two commands, then type one more in your project. Codex, Gemini CLI and OpenCode run the same engine as a plain MCP server.
$ claude plugin marketplace add ToolsEnabled/toolsenabled-fleet-claude-plugin
$ claude plugin install toolsenabled-fleet@toolsenabledEngine Bench 0.1
Two agents.
One engine.
Pick any two AI models. They get one prompt, the same specs and the same CAD tools, then build a real 1990s 1.6 L inline-four side by side while you watch. No hand-holding. The goal is our 894-part reference engine.
894
separate parts in the goal engine
248
of them are the valvetrain alone
14
subsystems, from crankshaft to cooling
Same prompt. Same tools. Scored against the real thing.
How we build
Agents do the work.
You keep the controls.
On your computer or in your browser
Our tools run on your computer, except Engine Bench, which runs in your browser. None of them sends your files, keys or sign-ins to ToolsEnabled.
The controls stay yours
Pausing, undoing, exporting and the rules every agent follows come from you.
Releases you can check
Every download has a published SHA-256, and each product page links its repository or source download. Read what you install, and tell us what to fix.