Evaluation
Free
Evaluate the complete governed workflow on lightweight tasks.
- Coroid-routed AI models
- 1 project
- 1 low-priority agent slot included
- 20 agent-hours included each month
- 2-hour maximum task run
- No parallel subagents
Transparent Pricing
See every part of the bill before you run a task: subscription, prepaid model usage, and any additional parallel capacity.
The $25 subscription is the base platform charge. Model usage and extra parallel capacity are separate, visible cost controls.
Parallel capacity
How many tasks should run at the same time. The base subscription includes one.
Model credits
Prepaid credits agents spend on model tokens. Add none and no paid tokens are consumed.
Your monthly bill
Provider token rates are passed through without markup. Credits roll over and are only spent as agents work.
Evaluation
Evaluate the complete governed workflow on lightweight tasks.
Base subscription
For organisations running a continuous backlog production line.
Contracted capacity
Contracted capacity, deployment controls, and support for larger organizations.
Everything you need to know about Coroid
Most general tasks are completed with about 100k–400k input tokens and 10k–40k output tokens. Depending on your AI profile, a typical task costs between $0.05 (Budget) and $1.50 (Smartest). See our token pricing table above for details.
You'll be prompted to top up. We never charge unexpectedly — you control your spending.
Every task flows through a 4-agent pipeline — Architect, Developer, QA, and Reviewer. Each agent validates the previous agent's work against the specification and acceptance criteria before a PR reaches you.
Yes! Enterprise customers can BYOK, and we're working on making this available to all paid tiers.
No. Coroid is entirely cloud-native. You write clear tasks in your browser, and the Dark Factory runs in isolated cloud environments. No IDE extensions, no local setup, no access to your organization's network.
No — and that's the point. Most AI assistants use a single model for all tasks. Coroid assigns a different model to each agent based on what it does best. The Architect runs a world-knowledge model for system design. The Developer uses a coding-specialized model. The QA agent employs a model proven at spotting edge cases and potential problems. The Reviewer uses yet another for static analysis. Each agent plays to its model's strengths, delivering far more consistent results than any single-model solution.
No, and here's why: we believe a sustainable business model is essential for long-term development. Your subscription directly funds improvements.