> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.fabricgateway.ai/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.fabricgateway.ai/_mcp/server.

# Know the current limitations

> Check the boundaries of Fabric AI Gateway before you commit, including deployment options, API translation, and access approval.

Knowing a product's boundaries saves a proof of concept. This topic lists current limitations that the rest of the docs describe, so that you can check them against your requirements. For hard numbers, such as request and body size limits, see [Limits and defaults](/reference/limits).

## Deployment

* **There's no fully self-managed option.** The control plane always runs in Postman cloud. On the Enterprise plan, you run only the data plane. For details, see [Choose a deployment model](/evaluate/deployment-models).
* **Some features need your own data plane.** JWT Auth validates tokens against an identity provider that you register with your data plane, so it isn't available when Postman hosts the data plane.
* **The Free plan leaves out some admin controls.** Granular RBAC, SSO, and admin controls are Enterprise features.

## Requests

* **The gateway doesn't translate between APIs.** It sends an OpenAI-format request only to a provider endpoint in the OpenAI format, and an Anthropic Messages request only to an Anthropic-format endpoint. A request to an endpoint that no provider on the route offers gets `404`. For details, see [Frequently asked questions](/help/faq).
* **A route uses one routing plugin.** The model router, the LLM Router, and the Adaptive Model Router can't share a route. To compare them, see [Choose a routing approach](/configure/routing/choose-an-approach).
* **The gateway doesn't redact PII from LLM responses or streams.** It removes PII only from the telemetry it records about them.

## Spend

* **Budgets don't notify you yet.** Fabric shows when spending nears or reaches a limit on the **Budget** page, but it doesn't send an alert. A budget set to allow requests at the limit only tracks spending.
* **Budgets are set for an organization or a namespace.** You can't yet set one for each end user from a request header. For details, see [Control spend with budgets](/configure/access-and-limits/budgets).

## Access

* **Callers can't approve their own access.** If access control refuses a request, the caller gets an error and must ask an admin. The gateway doesn't yet notify an admin or start an approval flow. For what the caller sees, see [Troubleshoot access](/use/troubleshoot-access).

## Product maturity

Fabric AI Gateway is still evolving, and it ships updates continuously instead of in numbered major versions. To follow what changes, see the changelog.

## Next steps

* [Limits and defaults](/reference/limits)
* [Try Fabric](/evaluate/try-fabric)