# Release 2026.8

Check August's updates.

***

<FeatureFlag flag="resources-skills">
## Skills for Your Agents

You can now give your Agents skills. Skills are versioned folders of files, such as instructions and reference material, that an Agent reads while it runs to work through a task the way you want.

On the new Skills page under *Resources*, you can browse skills made by deepset, create your own skills, and scope them to a workspace or your whole organization. You can also connect your organization's GitHub repository so its skills appear in the library. Once a skill is available, you can attach it to an Agent from the Agent's settings.

For details, see [Agent Skills](/docs/concepts/ai-agents/agent-skills.mdx) and [Manage Agent Skills](/docs/how-to-guides/building-agents/manage-agent-skills.mdx).
</FeatureFlag>

***

<FeatureFlag flag="services">
## Publish Your Pipelines Through Gateway

Your pipelines now have a stable home in production: Gateway. Inside Gateway, you publish a pipeline by creating a service, a fixed endpoint that never changes, even as you keep building and improving the pipeline behind it. Publish a new version whenever you're ready, and your callers stay connected the whole time.

- One endpoint, always. Iterate freely in the Builder without breaking what's live.
- Full history in one place. See every query, trace, and performance metric for your service.
- Hosting your way. Set replicas and idle timeout to match how much traffic your service needs.

We provide easy migration for pipelines you already deployed.

For details, see:
- [Gateway](/docs/concepts/gateway/gateway.mdx)
- [Publish a Pipeline to a Service](/docs/how-to-guides/productionizing-your-pipeline/working-with-services/publish-pipeline-to-service.mdx)
- [Manage a Service](/docs/how-to-guides/productionizing-your-pipeline/working-with-services/manage-services.mdx)
- [Migrate to Service](/docs/how-to-guides/productionizing-your-pipeline/working-with-services/migrate-to-service.mdx)
</FeatureFlag>

<FeatureFlag flag="haystack-version-headers">

## Pipeline Dependencies Block

You can now declare a specific Haystack Open Source version and add extra packages directly in your pipeline YAML using the `dependencies` key. When present, the platform automatically runs the pipeline in an isolated environment with exactly the packages you specify, with no changes needed to how you query the pipeline.

The platform validates the block when you save or deploy a pipeline, so version mismatches are caught early. For details on pipeline YAML, see [Building Pipelines](/docs/concepts/about-pipelines/about-pipelines.mdx#building-pipelines).

To learn more about Haystack Open Source and its relation to <ProductShortName />, see [Haystack Open Source](/docs/getting-started/whats-deepset-cloud/deepset-cloud-and-haystack.mdx).
</FeatureFlag>

<FeatureFlag flag="haystack-version-headers">
## Set the Haystack Version for a Pipeline 

Choose the Haystack version your pipeline runs on from the pipeline settings. Go to *Pipeline Settings* and select a version from the *Haystack Version* list. Selecting a version pins your pipeline to that release; leaving it unset keeps your pipeline on the default supported version. For details, see [Update Pipeline Settings](/docs/how-to-guides/designing-your-pipeline/update-pipeline-settings.mdx).
</FeatureFlag>

## Easier on the Eyes

You can now use the <ProductShortName /> in dark or light. Click your profile icon and choose your theme: Light, Dark, or System to match your system's settings. 

<ClickableImage src="/img/getting-started/theme.png" alt="Theme selector" size="large" />

<FeatureFlag flag="redesigned-conditional-router">

## Redesigned Conditional Router

The Conditional Router component has a redesigned drawer editor and canvas card. In the drawer, each route now appears as a sortable *If*, *Else if*, or *Else* block that you edit directly as YAML, with a Valid or Invalid badge that flags issues like malformed Jinja2 delimiters, output type mismatches, and duplicate output names for that specific route. On the canvas, each route gets its own condition row and a dedicated output handle, so it's easier to see how routes map to downstream connections. Renaming or removing a route automatically updates any canvas connections that depend on it.

To try this feature, ask deepset to enable it for your organization. For details, see [ConditionalRouter](/docs/reference/pipeline-components/logic-and-flow/ConditionalRouter.mdx).

</FeatureFlag>

## Self-Serve SSO Connections Management

Organization admins can now manage Single Sign-On (SSO) connections directly from *Settings > Organization > SSO*, without contacting support. Add a connection using a preset for your identity provider (Google Workspace, Okta, Generic OIDC, or Generic SAML), test it before saving, and edit, enable, disable, or delete it at any time. **SSO** is only visible to organization admins and is available on request. For details, see [Enable Single Sign-On (SSO)](/docs/how-to-guides/managing-access/enable-single-sign-on-sso.mdx).
