Use case
Claude Code across a whole team
Anthropic has published how its own teams use Claude Code: besides engineering, finance, growth marketing, product design and legal run their own workflows with it. Once a whole team uses it, usage comes from many people with different budgets. On BoostRail each person or department gets its own key with its own spend cap and model list, all drawing on one balance.
- Text models
- 61
- Spend caps per key
- Daily + monthly
- Model list per key
- Optional
- Organization request limit
- 600 / minute
What this workload needs
- Each person or department has its own budget, so one long session does not use up the shared balance.
- Staff outside engineering reach only the models chosen for them.
- Whoever pays can read usage by person or by department.
- Engineers and everyone else use the same Claude Code setup, with no second integration.
How BoostRail handles it
- One API key per person, department or pipeline; each key can have a daily and a monthly spend cap and a requests-per-minute limit, set in the console.
- Each key can carry a model list. Requests for other models return HTTP 403 (key_model_not_allowed) and are not charged, and GET /v1/models with that key lists only the models on it.
- Claude Code connects over /v1/messages with ANTHROPIC_BASE_URL and the key; Claude models and the other text models in the catalog run behind the same setup, and ANTHROPIC_MODEL picks the model.
- Call records in the console can be filtered by key, so usage reads per person or per department.
- All keys draw on the organization's one prepaid balance.
Set it up
- Decide the split: one key per engineer, one key per department where a team shares a workflow.
- Create the keys in the console, then set each key's spend caps and model list.
- Give each person the Claude Code config from the coding tools page, with their own key.
- After a week of use, read call records by key and adjust the caps.
Limits to know
Current behavior, the same as in the Docs.
- The console has no member accounts or roles yet: the account owner creates the keys and hands them out.
- Spend caps count completed usage, so requests in progress can take spend slightly past a cap. Requests over a cap return HTTP 429 until the next day or month (UTC).
- Models added to the catalog later are not added to existing model lists automatically.
- All keys in an organization share one limit of 600 requests per minute.
- Anthropic's documentation does not support running Claude Code on non-Claude models through another endpoint; test on your own codebase before relying on it.
Read next
Get your API key
Claude Code across a team: one key per person or department, each with its own spend cap and model list.
Last updated: 2026-11-12