About
An AI company that starts at the root.
Radicand Labs builds AI products for the people who build software. We design, build, run and support every one of them ourselves, starting with Afara.
(01)Our story
Our story
We're engineers and designers who kept losing time at the same seams between teams. Then coding agents arrived and the seams got wider: an agent only knows what's in front of it. We started Radicand Labs to build the tools that close that gap, beginning with Afara, and to build them the slow, careful way: problem first, evaluation second, product last.
(02)What we believe
Why we build what we build.
- 01
Agents need context
Coding agents are only as good as what they can see. The most valuable context lives in other teams: their code, their tickets, their decisions.
- 02
Trust is the product
An AI tool that engineers don’t trust is one they turn off. Privacy, predictability and plain explanations are features, not compliance.
- 03
Small teams, sharp tools
The best software of the next decade will be built by small teams with very good tools. We want to make those tools.
(03)Our products
One in beta, more in the lab.
Beta · Developer tools
Afara
Every team's context, in every coding agent. Afara means bridge in Yoruba.
How it worksIn development
Next
Our second product is being built the same way as the first. We'll share it when it passes its own evaluation.
Help build it(04)Principles
Six rules, written down.
Start at the root
Most AI products start with a model and go looking for a problem. We start with a problem we have felt ourselves, and only then ask what AI can do about it.
Private by design
We design every product so the sensitive work runs where the data already is. If a feature needs your source on our servers, we find another way to build it.
Measured, not vibed
Evaluation suites are the spec. Prompts, models and retrieval only change when the numbers say the change is better.
Inside the workflow
Our products live next to git, in the terminal and inside the coding agents teams already trust, not in another tab nobody opens.
Stable over clever
IDs don't get renumbered, accepted findings stay accepted, and a command behaves the same with or without a terminal.
Plain words
Error messages say what to run next. Findings say what the code does in a sentence. Documents read like a colleague wrote them.
Work with us
We're hiring across engineering, research and design.
See open rolesPress
For interviews, logos and product images, write to our press team.
press@radicandlab.co