Developer tools we wanted ourselves, most of them open source.
Products
Alongside our client work, we build developer tools for the platforms we work with every day — open-source command-line tools, and free ones that run in a browser.
A look at ebman — the kind of fast, keyboard-driven tool we build.
Capability maps and CI gates for AI-written code — candor maps which functions reach the network, the filesystem, a database or a subprocess, and fails the PR that crosses a declared boundary. Engines for the JVM, Rust, TypeScript/JavaScript and Swift.
brew install tombaldwin/tap/candor candor scan . candor mcp install
Install candor, map the repo (it fetches the engine it needs), then give your coding agent candor’s queries over MCP. On Homebrew 6 or newer, run brew trust tombaldwin/tap first.
A k9s-style terminal UI for AWS Elastic Beanstalk — triage red environments, stream logs and ship new versions, all from the keyboard.
or
On Homebrew 6 or newer, run brew trust tombaldwin/tap first. To try it without AWS, ebman --demo.
Optional — wire it into Claude Code:
A k9s-style Postgres TUI for Java and AWS shops. It auto-discovers datasources from your Spring config, turns Hibernate and RDS logs (and pasted JDBC) back into runnable SQL with N+1 detection, and adds safe-by-default DBA panels — schema browser, EXPLAIN, slow queries, locks and more.
Paste any URL and see how it unfurls on X, Facebook, LinkedIn, WhatsApp and Slack — with a checklist of what’s missing or off. There’s a JSON API for scripts and agents too.
Keeps watch on a site’s social cards. goosey remembers every page it checks, in your browser, says what changed when you check again, and gives your build a check that fails when a card breaks. One check reads a sitemap and plans up to 40 pages across its templates, and nothing more is spent until you agree.
Finds the Carnot limit of an AI coding agent fleet: how many agents to run, what is holding them back, and how long the work will take. An Agent Skill: ask your coding agent “how many agents should we run on this repo?” and it measures the repository’s own history and runs the model from The Heat Death of the Codebase?
/plugin marketplace add tombaldwin/carnot /plugin install carnot@carnot
In Claude Code. For other agents, copy the skill folder from the repository. Python 3.9+ and git; MIT licence.
Questions about one of these tools, or a system of your own to build?
Get in touch