Problem The current BYOM configuration accepts an API key for a supported provider (Anthropic, OpenAI, Gemini), but does not allow specifying a custom base URL. Enterprise organizations typically route LLM traffic through an internally-hosted proxy or gateway (e.g., LiteLLM) to enforce security controls, model governance, and cost management. In these environments, using a personal or team API key tied to a commercial provider account is not viable. All traffic must go through an internal endpoint on an approved FQDN. Requested Solution Add a custom base URL field to the Spacelift AI BYOM configuration, alongside the existing API key field. This would allow Spacelift to direct AI requests to any OpenAI-compatible endpoint (e.g., https://llm.internal.example.com/v1) rather than only to a commercial provider's public API. This pattern is standard across OpenAI-compatible clients (OPENAI_BASE_URL, openai.base_url in the Python SDK, etc.) and is how tools like LiteLLM, Azure OpenAI, vLLM, and Ollama are accessed. Use Case Our organization runs a centrally-managed LiteLLM instance serving internally-approved models via an OpenAI-compatible API. We want Spacelift AI features (plan summaries, resource explanations, policy suggestions) to use this internal endpoint, but the current BYOM flow only accepts a provider API key, with no way to redirect the base URL. Priority High. This is a blocker for enterprise customers with internal model governance requirements.

You are unable to create stack dependencies without space admin

When creating a run, now we have the option to force apply that single stack or in cascade. The issue surfaces when the run is queued or ready, because we cannot know that that option was enabled. It is only shown after the plan phase, because it shows the apply phase with the force applied. I think it would be very useful and interesting to have the ability to see that somewhere in the stack of runs and also in each individual run. Together with this, I think that should be added to the GraphQL API for when listing the runs for the workers queue, and add the filter in the workers queue UI.

As an example, when iterating on login policies, every change invalidates all active sessions for the affected account. This creates a feedback loop: authors must repeatedly log back in (and disrupt other users' sessions) each time they test a policy modification, even for minor fixes like correcting a role name. The existing policy simulator helps validate syntax and basic logic, but it does not catch real-world authorization issues. For example, a policy may reference a role name like writer instead of the correct space-writer. The simulator evaluates the policy as valid, but the role doesn't actually grant the intended permissions in practice. These kinds of bugs are only discoverable through live testing — which currently means deploying to production and invalidating sessions.

When using OIDC API keys for authentication, JWT teams/groups claims are completely ignored, only the sub claim is processed. Teams must be pre-configured statically when creating the OIDC API key, making it impossible to pass through user group/team information dynamically at runtime. We are building a custom Backstage integration with Spacelift to enable self-service infrastructure provisioning with per-user permission boundaries. The Backstage plugin is not suitable as it uses a single admin API key. We need OIDC API keys to pass through the current user and respect Spacelift login policies, as is currently possible with SAML via input.session.teams. With thousands of team/service combinations, the existing workarounds (static API keys per team or subject-based encoding) are not viable at scale.

We sometimes get PRs which affect many different Stacks. Because each Stack posts an individual comment with the proposed run status, the PR conversation section can become bloated and basically unusable. Also, we have hit GitHub API rate limits and the comments may have contributed here. We would like there to be an option to merge proposed runs into a single PR comment which would gather all proposed runs for given PR. We keep comment content quite short, so the GitHub comment character limit should not be an issue.

It looks like the current tools in the MCP server only allow either getting a specific run using a stack id and run id, or listing the runs in a stack. For some use cases it would be more efficient to have a tool that allows searching or filtering the runs. In particular so that an agent can find the proposed runs for a pull request based on the commit id

When a Terraform module includes an examples/ folder, the Spacelift Module Registry only shows a README (if one exists) under the Examples tab. There is no display of the actual.tf files (e.g. main.tf) that contain the example configuration. This mirrors a known limitation in the public Terraform Registry. The Examples tab in the Spacelift Module Registry renders only the README for a given example subfolder. If no README exists, nothing is displayed at all. The actual.tf files are not surfaced anywhere in the UI. The.tf files within an examples/ / subfolder should be displayed directly in the UI — either alongside or as a fallback to the README — so users can understand how to use the module without navigating to the source code repository. Looking to render.tf files from the examples/ / directory in a syntax-highlighted code view within the Examples tab, either as additional tabs per file or as a collapsible file tree below the README.

The current Pending runs count is misleading because it lumps together high-priority and low-priority (e.g. drift detection) runs without any distinction. Users can see a large pending number vs. a small worker count and panic. Surface run priority/ordering in the queue UI Show expected wait time or at least a priority-aware count something like: Workers: 6  Busy: 6  Pending: 7 (+ 183 low-priority tasks) as a quick win, with a long term goal of an AI-estimated wait time in minutes.

you can click on Custom Roles, in which you can view two tabs (Permissions and Used By) For the System roles (default reader, writer, admin) you cannot click on them. It makes sense you cannot configure the permissions but I want to be able to view the “Used By” tab

It would be helpful to set a private workerpool as the default for all stacks organization-wide.

State import is only available through the Spacelift UI. Users must manually click through each stack one by one to import state files. For customers migrating large numbers of stacks, this is a significant manual bottleneck. Desired behavior A CLI command (e.g. spacectl stack state import) that accepts a state file path and a stack identifier, allowing state import to be scripted and automated as part of a migration pipeline.

Templates currently resolve the VCS repository field at publish time, meaning the repository is baked into the template version and can't be provided as a dynamic input. This prevents using Templates for use cases where the same stack configuration needs to be deployed across multiple repositories - e.g. a "Stack Vendor" template that engineers deploy from to onboard their repo.

Please implement RBAC for source code integrations. We do not allow our user-base to assume ‘space admin’ privileges. We make the most of the new custom roles feature to allow customer teams to create their own stacks. However, we have to manage their source code integrations on their behalf as there are no available RBAC actions that can be assigned to custom roles to create Github / Gitlab / Bitbucket integrations. Could you please implement this? It would vastly simplify the administrative overhead. Many thanks and loving your work.

Add data sources for working with templates in the Terraform provider, such as spacelift_templates to list templates and spacelift_template_version_by_name to resolve a version by name (for example 1.0.0).

When a run fails unexpectedly, stale Terraform state locks are sometimes left behind. Request to have this condition detected and automatically clean up locks that were created as part of a run.

We would like the ability to share private worker pools across spaces without requiring space inheritance to be enabled. Ideally, this would support sharing between sibling spaces, in a similar way to how modules can currently be shared.

I would like to be able to systematically monitor our worker pool to get quantitative data on our internal developer experience as our engineers all share the worker pool to run their IaC. I would like to use those metrics to improve decision making on the number of workers we need for our organizations needs.

“Mark version as bad” is informational only. Enterprise customers need a structured lifecycle so module authors can deprecate versions with a grace period and then block unsupported versions without maintaining brittle external OPA logic. In many organizations, infrastructure patterns are encapsulated into approved modules (e.g., networking, S3 buckets, IAM roles, etc.). When these patterns are modularized, developers can safely self-serve infrastructure by consuming those modules rather than building resources directly. Proposed solution Add first-class lifecycle state for each module version: active deprecated (optional sunset_date) unsupported Ideal behavior Using a deprecated version: plan succeeds but emits a warning stating it’s deprecated, recommended version, and sunset date (if set). Using an unsupported version: plan fails (hard stop). Acceptance criteria Registry UI shows lifecycle state per version and (if applicable) sunset date. Lifecycle state is persisted per version and queryable via GraphQL. Deprecated usage generates a warning surfaced in the run. Unsupported usage blocks the run automatically.