Why Arkoria
Others charge for their cost. We charge for your value.
Most suppliers bill for what the work costs them. Arkoria bills only for stories you approved, once they are delivered and verified.
- Per hourmeasures the vendor's effort
- Per seatmeasures headcount
- Per token or creditmeasures model consumption
- Per generated linemeasures output volume
- Per approved storymeasures work you approved, delivered
Pricing, from first principles
Every pricing unit is an incentive.
When the bill grows with hours or consumption, rework is revenue. When it grows with approved, verified work, rework becomes the supplier's cost. We chose the second on purpose.
| Pricing unit | What it measures | Who pays for rework | What it rewards the vendor for |
|---|---|---|---|
| How delivery vendors charge | |||
| Hours | The vendor's time | You: rework is more hours | More hours |
| Fixed scope | An estimate made before the work was understood | You, once rework becomes a change request | Change requests |
| Dedicated teams | Headcount assigned to your account | You | Bigger teams |
| How AI tools charge | |||
| Seats | Headcount with access | You | More seats |
| Tokens and credits | Model consumption | You: retries and rework consume more | More consumption |
| Compute time | How long agents ran | You: longer sessions cost more | Longer-running work |
| Cost-plus | The vendor's model cost, plus a margin | You: the margin grows with the cost | Higher underlying spend |
| Generated lines of code | Output volume | You: every line costs, accepted or not | More code, whether or not it is better |
| How Arkoria charges | |||
| Approved stories Arkoria | Work your people approved, delivered and verified | Arkoria: rework stays inside the story, and bug stories are never billable | Delivering, and finding defects first |
The market is moving the same way. DORA's 2026 guide to the ROI of AI-assisted development concludes that governance, not inference, is now the main cost of adoption: raw inference prices for advanced models fell roughly 280-fold between late 2022 and late 2024, per the Stanford AI Index. DORA, The ROI of AI-assisted Software Development, 2026
Incentives, aligned
Built so our interests point the same way as yours.
-
01
Quality costs you nothing
Bug stories are never billable, and the platform opens its own. We are never paid to create a defect and then paid again to fix it.
-
02
A neutral referee
Arkoria sells no agent and no tokens. It has no reason to make you consume more, and every reason to make sure what you approve gets delivered.
-
03
Rework is ours
A story blocked at review goes back to the agents without a second line on your invoice. It is still one story.
-
04
The price falls as you take over
$85 a story when we deliver, $70 when your team builds on our inference, $60 on yours. Building your own capability lowers the price. There is no penalty for needing us less.
-
05
Your work stays yours
Your code, backlog and project documentation are yours, and the documentation is written to outlive the vendor relationship.
-
06
Nothing to unwind
Onboarding never modifies your source, CI, database or dependencies. Arkoria's governance files sit alongside yours, so if you stop there is nothing of yours to put back.
What changes with Arkoria
Agent platforms run agents. Arkoria governs what they ship.
Coding agents, review bots and agent platforms each cover part of the job. None of them can tell an auditor, a client or your board who authorized a change and what it had to pass before it shipped, because that takes the whole delivery process, with everyone's part in it recorded.
| Question | Coding agent alone | Agent + CI and a review bot | Agent + Arkoria |
|---|---|---|---|
| Who it works for | The developer at the keyboard | The developer and the reviewer | Everyone with a say: the people who ask, approve, build and accept |
| Scope before code | Whatever the prompt said | Whatever the prompt said | A spec a person approved, with no code before it |
| Tests before code | If the agent follows instructions | CI checks tests pass at merge; nothing checks a test existed and failed before the code | A failing test is required before code may be written |
| Review | A person reads every diff | Required reviewers plus advisory bot comments, on the pull request | A fresh-context review whose critical findings block before anything ships |
| “Done” means | The agent says so | Required checks pass and someone merges | After the merge, the platform verifies the commit on the target branch rather than taking the agent's report |
| Human decisions | Scattered across chats and pull requests | At the pull request | Five, named: request, charter, spec, governance change, acceptance |
| Evidence left behind | A transcript and a diff | Plus review comments | A record that can only be added to: consent, evidence and review |
| Billing unit | Seats, tokens or credits | Plus seats or credits | Approved stories; bugs never billable |
Arkoria governs the delivery process and works alongside all four of the categories below.
Where the other categories focus
- Coding agents write the code and govern their own sessions.
- Agent platforms govern what agents may access and spend: sandboxes, secrets, budgets.
- AI code review comments on one pull request at a time.
- Engineering analytics reports on delivery after the fact: it measures, it doesn't block.
Build, buy, own
Free to run isn't free to trust.
Open-source agent platforms give you workflows and approval gates for free. What takes years is making the output trustworthy, with rules agents can't get around, proof that each commit reached your codebase, a reviewer that can block the work, and drift you can measure.
- Replaced a SaaS tool with a custom build
- 35%
- Built software outside IT oversight
- 60%
Retool, Feb 2026 · 817 respondents · vendor sample
Who edits the rules?
In Arkoria, governance changes only happen inside a governed story, approved by a person.
The path ends with your team in control: pilot, project, product
Questions buyers ask
The hard ones, answered plainly.
We already use a coding assistant. Isn't this the same thing?
An assistant makes each developer faster at writing code. It does nothing to settle what should be built, check the result independently or leave you with evidence, so the work piles up at review, testing and sign-off. Arkoria runs that whole path, from the request to a shipped, proven change your people test and accept.
Is this a headcount-reduction play?
No. Judgment is the part of an engineer's job that matters, and it stays with your engineers. They write the specifications, approve the plans and design the rules the code is written under. What you gain is capacity for the roadmap you had to cut, with the team you already have.
Can work done by non-deterministic agents be audited?
The process can, and the process is what an auditor samples. No code is written before a person approves the plan. Every gate produces a structured verdict tied to the story. Approvals, evidence and reviews are recorded in order, in a record that can only be added to.
Does our source code leave our infrastructure?
It doesn't have to: the agents that touch your source run on your infrastructure or ours, confirmed in writing for each account, and the hub reads your repository through GitHub's API, opening and merging pull requests only on projects that use pull-request review. Where it runs, and what it touches.
Are we locked to one AI vendor?
No. The rules are held in Arkoria's hub rather than in any one agent, so regardless of which agent wrote the code, no story closes without its approvals and a verified commit. It runs natively on Claude Code today, with OpenAI Codex coming soon; which agents qualify.
What if Arkoria disappears?
You keep the code, the backlog and the documentation, which is written so the work can carry on without us. Continuity and escrow terms can be negotiated in the contract.
The name
ἀρχή
arche · origin, and authority
Origin and authority, in one word.
Arkoria takes its name from the Greek arche, a word that means both origin and authority, and which the earliest philosophers used for the first principle from which everything else follows. It is the root of archon, the magistrate who governed, and of architect, the chief builder.
That double meaning is the product, and the reason our line is Governance from first principles. The approved specification is the origin and the authority: every artifact downstream must prove it descends from that decision.
In Portuguese, the ending lands in the vocabulary of oversight (auditoria, corregedoria, ouvidoria), and Arkor·IA carries the letters for artificial intelligence. Say it ar-KOR-ee-a.
Start small
Start with the project you keep putting off.
We'll scope a first piece of it as a Governed Pilot, built and shipped on your own codebase at a fixed price.
- Duration
- 4 weeks from requirements
- Price
- $3,500 fixed, half on signature and half on delivery
- Delivered
- One epic of up to 10 stories, shipped with every approval and review recorded
- You keep
- The code, the documentation, the planning report, a backlog of small, testable stories and a governance report
- Scope
- One workload, up to five users; cutover, migration and certification out of scope
- If you stop
- Nothing to unwind, because your source, CI, database and dependencies were never modified