Put the capacity you already pay for to work.
Desktop agents run on subscriptions you have. Cloud agents run the best agent on a model picked for the task, inside spend limits.
Use the tokens you already pay for
- Individual
- Enterprise
- API
Excess capacity
Desktop agents use the capacity in your subscriptions that would go unused.
A more engaged team
More developers take part in the process, everyone is better trained, and work stops going to the same few people.
Higher productivity
Idle capacity becomes finished work.
The best agent with an efficient model
Claude Code
Claude Code runs the cloud task.
Claude
Picked for the task
Match each task to a model for quality and cost. Open-source models are a lower-cost choice where they fit.
Hosted models
Claude
GPT
Gemini
Open-source models
Qwen
DeepSeek
Kimi
Choose the model for each task
Cheapest, balanced or best
Set a preference. Cortega picks a model for each task from the models the agent can use.
Per task
A small fix and a large refactor do not need the same model.
Your agent's own models
The pick is always one of the models the person's agent reported, so it can run.
Spend limits at every level
Cost cap per task
A cloud task stops when it reaches its cap. The default is $5.
Monthly limit per cloud agent
Each cloud agent's gateway key has a monthly spend limit. The default is $50.
Company budget
A company budget sits above the keys.
Usage cap per agent
Each desktop agent has a usage cap and a number of slots.
Spread across the team
Lower recent usage ranks first, so work spreads across the team.
Cost on every run
A cloud run reports its model, tokens and cost.