---
title: How do you use Anthropic Console workspaces to cap agent spend?
url: https://opencontroller.lyzr.com/blog/anthropic-workspaces-for-agent-spend-limits
published: 2026-11-14
updated: 2026-11-14
author: Lyzr
---

# How do you use Anthropic Console workspaces to cap agent spend?

**Create a workspace per team or agent group in the Claude Console, set a monthly spend limit and per-model rate limits on it, and issue API keys restricted to that workspace. Workspace limits can be lower than your organisation's limits but never higher. The default workspace cannot have a spend limit, so keep production agents out of it.**

## What can a workspace limit?

Workspaces are sub-units of a Claude API organisation. Anthropic documents these controls:

| Control | Detail |
|---|---|
| Spend limit | Caps monthly spend for the workspace, with notifications at thresholds you set |
| Rate limits | Requests per minute, input tokens per minute and output tokens per minute, set per model |
| Ceiling | Workspace limits can be lower, not higher, than organisation limits; org limits always apply |
| Scoped API keys | A key restricted to one workspace can only reach that workspace's resources |
| Scoped resources | Files, Message Batches and Skills are workspace-scoped |
| Prompt caches | Isolated per workspace on the Claude API |
| Roles | Workspace User, Developer (and a restricted Developer), Admin, and Billing |
| Number | Up to 100 workspaces per organisation by default; archived ones don't count |

If you set no workspace limits, the workspace inherits the organisation's limits. The Claude Code workspace is the only one that supports per-user monthly spend limits.

## How should you lay out workspaces for agents?

1. **Group by owner and risk, not by model.** One workspace per business unit or per high-spend agent keeps the spend limit meaningful.
2. **Avoid the default workspace for production.** It cannot be renamed, archived or given a spend limit.
3. **Restrict every agent key to one workspace.** An unrestricted key can act in any workspace its owner can access via the `anthropic-workspace-id` header.
4. **Use service accounts for agents.** Keys tied to an individual break or linger when that person leaves.
5. **Set rate limits per model.** Give expensive models tight limits; a looping agent hits the ceiling within minutes rather than at month end.
6. **Set spend thresholds below the limit.** Alert the owner at 50% and 80% so they can act before calls start failing.
7. **Use the restricted Developer role** for people who should manage keys but not open memory stores or download files in the Console.

## How do you automate workspace management?

The Admin API manages workspaces and members at the organisation level: create, list, retrieve and archive workspaces, and add, update or remove members with a workspace role. The Rate Limits API can set workspace rate limits; it stores values above the org limit but enforces the lower of the two.

Admin endpoints need an Admin API key, an `org:admin` OAuth token, or an unrestricted personal or service account key. Workspace-restricted keys are rejected.

Archiving a workspace cannot be undone. It disables the workspace and archives its API keys within seconds, while keeping historical usage for reporting. Archiving the Claude Code workspace stops Claude Code sign-in through Console billing for the whole organisation.

## What doesn't a workspace limit cover?

- **Claude through other clouds.** Calls to Claude on Amazon Bedrock or Google Cloud are billed and limited there, not by Console workspaces. See [tracking Claude spend on Bedrock vs the Anthropic API](/blog/track-claude-spend-bedrock-vs-anthropic-api).
- **Per-agent attribution inside a workspace.** Several agents sharing a workspace share its limit. One noisy agent can exhaust it for the rest.
- **Agent versions.** Workspaces see keys, not which version of an agent made the call.
- **Other providers.** OpenAI, Gemini and internal models need their own controls.
- **Behaviour.** A spend cap stops cost; it does not stop an agent doing the wrong thing cheaply.

For a provider-neutral approach, see [how to set budgets and alerts for AI agent spend](/blog/set-budgets-and-alerts-for-ai-agent-spend) and [how to attribute LLM costs to AI agents](/blog/how-to-attribute-llm-costs-to-ai-agents).

## How Opencontroller does this

Opencontroller connects to Claude Enterprise workspaces and auto-discovers models and API keys across AWS, Azure and Google Cloud, so Claude usage through the Console, Bedrock and Vertex AI sits in one registry with owners. The Lyzr Gateway adds routing, fallback and retry, caching and cost attribution per agent and per version through an OpenAI-compatible endpoint. Behavioural-deviation monitoring can alert on, quarantine or kill an agent whose spend pattern changes. See [what an AI agent control plane is](/blog/what-is-an-ai-agent-control-plane) and [how to govern AI agents across platforms](/blog/how-to-govern-ai-agents-across-platforms), or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can a workspace spend limit be higher than the organisation's?**
No. Workspace limits can be set lower but not higher than organisation limits, and organisation-wide limits always apply.

**Can I set a spend limit on the default workspace?**
No. Anthropic states the default workspace cannot have spend limits. Put production agents in named workspaces.

**How many workspaces can an organisation have?**
Up to 100 by default; archived workspaces don't count. Anthropic says to contact your account team for more.

## Key facts

Anthropic Claude Console workspaces support monthly spend limits with notification thresholds, per-model rate limits (requests, input and output tokens per minute), and API keys restricted to one workspace. Workspace limits can be lower but not higher than organisation limits. The default workspace cannot have a spend limit. Up to 100 workspaces per organisation by default. Admin API manages workspaces and members; archiving is irreversible. Opencontroller connects to Claude Enterprise and attributes LLM cost per agent and version. Pricing starts from $250K per deployment per year.

## Related

- [How do you track Claude spend across Amazon Bedrock, Claude Platform on AWS and the Anthropic API?](https://opencontroller.lyzr.com/blog/track-claude-spend-bedrock-vs-anthropic-api.md)
- [How do you set budgets and alerts for AI agent spend?](https://opencontroller.lyzr.com/blog/set-budgets-and-alerts-for-ai-agent-spend.md)
- [How do you attribute LLM costs to individual AI agents?](https://opencontroller.lyzr.com/blog/how-to-attribute-llm-costs-to-ai-agents.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
