Home›Companies›Rystic
Rystic

Test agents and code changes in high-fidelity simulated environments

Staging environments and test sandboxes don't scale for development with AI. Instead we give every change a deterministic simulated environment populated with local, stateful simulators of APIs like Slack, Stripe, GitHub, and PostHog. Control everything about your dependencies, from the clock to injected faults, and use our harness for integration tests or agent evals. With simulation testing, every change is evaluated quickly in its own private world where you can measure its impact on other systems. We're building the next step forward in software testing, one that scales with your agents instead of holding them back.
Active Founders
Evan Whitehouse
Evan Whitehouse
Founder/CEO
Founder at Rystic. Networking Infra at Roblox; Built and scaled the service mesh to support 45 million concurrent players. Columbia CS grad. Former professional Magic: the Gathering player.
Zachary Kiihne
Zachary Kiihne
Founder/CTO
Founder at Rystic (YC F26). Previously Senior Software Engineer at Google, where he shipped Gemini AI features in Gmail including conversation UI and email summaries. Former professional Magic: the Gathering player, ranking 2nd in the world in 2022.
Company Launches
Rystic: Test your AI agents with local copies of Slack, Stripe, and Google Drive 🧪
See original launch post

Hey everyone, we’re Evan Whitehouse and Zachary Kiihne from Rystic 👋.

TL;DR

Rystic gives your agent working copies of the tools it acts on: Slack, Stripe, Google Drive, and 14 more. Your tests can check what the agent actually did, not just what it said. Each copy keeps its state, runs locally or in CI, and is checked against the real service every day .

Every test job gets its own isolated copy, so tests run in parallel: 50 Stripe test jobs finished in 19 seconds, compared with 40+ minutes on shared real-API sandboxes.

uploaded image

Asks: Building an agent that acts on real systems? Try free on rystic.ai. For custom environments, including internal services, get in touch at founders@rystic.ai.

The Problem

At a previous job, the first agent we gave write access to GitHub flooded our CI runners by creating hundreds of PRs within a few minutes. There was no way to iterate on our agent without running it against our production systems.

When giving an agent write access to a system with real consequences, how can you trust that they’ll behave properly?

https://youtu.be/8dvzTRv_la8

How We Solve It

Rystic can simulate an API or Web app as a lightweight binary. Run it anywhere and get behavior that matches the real thing. These simulators are stateful: delete data, query it, and you’ll get the same error the real service returns.

A simulator also lets you test things you never could with the real API. What happens when latency triples? What if the data feed’s websocket connection drops in the middle of placing an order? With a simulator, you can script the failure cases that matter and prove you handle them correctly.

We use these to create an environment which can be seeded with either real trace data or synthetic data, so you can validate your agents’ actions and their impacts before you release them.

Why You Can Trust It

How do you know the simulator behaves like the real thing? We check every simulator against the real service daily and publish the results. Our simulator matches Stripe on 99.1% of 1,813 checks, and every known difference is listed.

uploaded image

Catalog

uploaded image

The Team
We met all the way back in 2016 when we were up and coming professional Magic: the Gathering players. We practiced together and spent over a year on the same pro team. Evan finished 2018 ranked #2 on the SCG tour, the largest independent circuit; Zach made the Top 8 at three Pro Tours, the game's premier event. Then we both pivoted into software engineering.

Since then, Evan graduated with honors from Columbia CS and helped build Roblox’s service mesh from 0→1, supporting hundreds of millions of concurrent requests. Zach has spent 5 years at Google, and shipped AI integrations in Gmail and Google Docs to billions of people.

We learned to win at Magic by finding the edge cases everyone else missed, so we’re building Rystic the same way.

Rystic
Founded:2026
Batch:Fall 2026
Team Size:2
Status:
Active
Location:San Francisco
Primary Partner:Grey Baker