One agent per domain, two domains so far
The top layer of the stack: independent agents, each built for one professional domain and run by its own harness. Today there are two: Runix Code for software engineering and Runix Comic for media and entertainment, both in development with waitlists open. The layer is presented the way a buyer thinks, by the work to be done.
Two agents exist today, both in development with waitlists open. The rest of this page says where the next ones come from, without promising one.
The layer at a glance
- Software engineering
- Runix Code · in development
- Media and entertainment
- Runix Comic · in development
- Runs as
- An independent agent per domain, each with its own harness
- Gateway
- Runix Router is optional in front of an agent's model calls
- Candidates
- Finance, Cybersecurity, Legal, AI for Science: domains the Data layer already has rules for
02 · Definitions
What an agent in this layer is
Three definitions the rest of the page depends on.
What a domain agent is
A domain agent does the work of one profession, inside that profession's material: a coding agent in a repository, a studio agent in a script and its episodes. It is independent: it runs on its own and does not depend on the rest of the stack to do its job. What it produces is something a person in that domain already knows how to review.
What the harness is
The harness is what runs the agent: the loop that takes an instruction, plans the work, calls the model, applies the result to the material and stops where a person has to look. Each agent has its own, built for its domain, because the steps that make sense in a repository are not the steps that make sense in an episode.
Where the gateway sits
Runix Router is optional in front of an agent's model calls: one OpenAI-compatible endpoint, with keys, quotas and allowed models beneath it. The agent runs independently of it. Put Router in front when you want the agent's model usage governed and billed the same way as every other call.
03 · By domain
The agents, by professional domain
One agent per domain, each built around the tools, data rules, evaluation and review gate of that profession.
For engineering teams
Software engineering
Served today by Runix Code · An independent coding agent, reviewable in your repository
Runix Code is an AI coding agent being built for teams: grounded in your repository, with multi-file implement, refactor and test from one instruction, and every change landing as a reviewable diff. Seats, per-key limits and usage attribution by developer are part of it. In development, waitlist open.
Status: in developmentFor studios and creators
Media and entertainment
Served today by Runix Comic · An independent agent for comic dramas, script to episode
Runix Comic is an AI studio agent for comic dramas being built around the whole job: scripts, episode plans, consistent characters and storyboards; styled panels, motion and camera, voice, music and subtitles; vertical cuts and full episodes for the platforms you name. In development, waitlist open.
Status: in development04 · Next
Where the next agents come from
An agent needs its domain's rules before it needs a model. The Data layer already has rules for these four; none of them has an agent announced, and this page does not announce one.
05 · Principles
What every agent here must satisfy
Reviewable output
An agent's output is something a person in the domain reviews before it counts. Runix Code is built so every change lands as a reviewable diff, never a silent write; the same rule holds for every agent in the layer.
Grounded in your material
Runix Code works from your repository: its modules, idioms and the versions you run. Runix Comic works from your script, with characters that hold across scenes; neither starts from a tutorial's example.
Team controls
Seats, per-key limits and usage attribution by developer, as Runix Code is being built with, so an organisation can roll an agent out rather than discover it. The controls are set up with you, not left to the honour system.
Data terms in writing
Where Runix Router sits in front of an agent's model calls, its written data terms apply: content processed to serve the request, never used for training. The contracting entity is Runix AI Inc, with an MSA and a DPA on request.
06 · Questions
Common questions
How many agents exist today?
Two: Runix Code for software engineering and Runix Comic for comic dramas. Both are in development and not generally available; their waitlists are open.
Does an agent need Runix Router?
No. Each agent runs independently, with its own harness. Router is optional in front of its model calls, for teams that want one endpoint, per-key limits and one itemised bill.
Will there be an agent for my domain?
Nothing beyond Code and Comic is announced, and no dates are published. The Data layer's six domains are where the domain judgement already exists, and waitlist cohorts set the order of what gets built.
Tell us which domain you work in
The job you want an agent to take over, the material it would work on, and who reviews its output. We reply within one business day with where the nearest agent is and what it can do today.
Contact sales