Empathic

Agency in a world
full of agents

Enforceable control, real-time collaboration, and full provenance for AI agent infrastructure. Built for teams crafting for the future.

Request Access >>
The Problem

AI agents are rapidly defining a new world of software engineering. This is a double-edged sword.

Our legacy software practices are breaking down

We're all building more, faster. The upside is that more bugs from our backlog are getting knocked out and more features launch every week than ever before!

The downside is that in a given week, we're introducing a month's worth of code without the practices and tools to keep up. Teams a year into this journey are inventing their own specialized systems to not be buried under mountains of technical debt.

We need the tools to work better, together, as humans

Using AI agents today is like playing slots. We jack in and are absorbed into our glowing screens, get the dopamine hit of results spinning by, and feel productive -- but it's lonely, isolating, and disconnected from the people around us. While we're individually more productive building software, slinging 5000 line PRs is a frustrating way to work together.

We're trusting the untrustworthy

These agents we're all trusting to extend ourselves into this future are executing with the guardrails intended for us, but these agents aren't us.

Context degrades silently. Rules are advisory, not enforceable. The more powerful the models get, the wider the gap between what agents can do and what humans can control.

Our Products
[01]

Clash

Control & Trust through Agentic Sandboxing

A majority of Claude interruptions are asking whether it's okay to run some command. This is why --dangerously-skip-permissions is so powerful: we can achieve so much more per unit prompt.

But why is that the choice? Why must we choose between becoming a pedantic babysitter or giving Claude's actions dangerous free rein over our system?

Clash gives us the best of both worlds. It allows sandboxing what matters: the actions Claude is performing. Every tool, every command, independently.

The Clash plugin for Claude runs in two phases,

  1. matching and deciding what actions Claude can perform, and
  2. dynamically selecting the sandbox in which that action will be performed.

Unlike wrapping our agent in a container or VM, we are in control of which tool gets what access.

With Clash, we get the productivity of --dangerously-skip-permissions without losing the control.

Contact

Built for teams who care

Empathic is building open tools to define the new craft of engineering. Join us in increasing human agency in a world full of agents.

Request Access

Empathic is a small team in Flatiron, NYC. Engineers from Meta and Google who've worked on enterprise AI infrastructure, human-centered product, and pragmatic distributed systems. We build tools for the human-agent relationship: control, collaboration, transparency.