MCP Tools Reference

Overview

When you connect an AI assistant to the StackGuardian MCP server, it gets a set of tools. A tool is one specific action the assistant can take in StackGuardian on your behalf, such as searching your workflows, fetching a run's logs, or starting a plan.

You never call these tools yourself. You ask your question in plain language, and the assistant picks the tools it needs, often chaining several together. For example, "Why did the last run of payments-api fail?" usually becomes:

1. search_workflows to find the payments-api workflow
2. search_workflow_runs to find its most recent run
3. get_workflow_run_logs with only the error lines

This reference explains what each tool does, so you know what you can ask for and what happens when you do.

How the tools work

Your organization comes from the credentials you connected with, never from anything the assistant types. Tools use it to limit what they return to your organization and to what your role is allowed to see, going through the same permission checks as the platform itself.

Tools that act on a specific workflow, run, template or connector expect its real id. Assistants are instructed to look ids up with a search tool first, instead of guessing or asking you to type them from memory. When a name doesn't match, most tools return the valid options so the assistant can offer you a choice.

Specific tools protect sensitive data in their responses:

• Workflow outputs marked as sensitive are redacted by get_workflow_outputs.
• Artifact and Terraform state tools share short-lived download or upload links for you to open, rather than file contents.
• Vault-secret tools return references and metadata, never the stored secret value.

These protections don't cover every kind of tool result. Console logs are returned as text without an additional redaction pass. If a workflow prints credentials or other sensitive values, asking the assistant to fetch those logs can put them into the conversation. Plain-text values in resource configurations can also appear in tool results. Anything you type into the chat yourself, such as a new secret value you ask your assistant to store, is part of your conversation too.

Read tools and write tools

Tools come in two kinds:

• Read tools look things up or present results and change nothing in the platform.
• Write tools create, change or delete things in StackGuardian, and some run infrastructure code. Each one has safeguards built in (see below).

Availability

Write tools are enabled per environment. If your assistant doesn't list a write tool, it isn't enabled for your organization yet. The read tools are always available.

Safeguards on write tools

Write tools are designed to keep you in control:

• You choose the action. Tools that run or control infrastructure have no default action. The assistant is instructed to ask which action you want (plan, apply, approve, cancel and so on) and wait for your answer.
• Explicit confirmation. Irreversible operations, such as destroy runs, deleting a workflow or secret, approving a run or overwriting Terraform state, require an explicit confirmation flag. The assistant should only set it after you've said yes.
• Validated before sending. Inputs are checked before anything reaches the platform. A mistake comes back as a plain explanation, often with the valid choices listed, instead of an opaque API error.
• Read before write. Update tools read what's stored first and change only the settings you name. Some settings replace the stored value instead of merging into it; each tool's page lists which, so check it before asking to add to a list.

tip

Install the StackGuardian agent skills alongside the MCP server. Skills walk your assistant through multi-step tasks, such as creating, updating or upgrading a workflow, in a consistent, reviewed order.

All tools

Cloud posture

Headline numbers and breakdowns for your cloud resources: IaC coverage, drift and misconfigurations. See Cloud posture tools.

Tool	What it does	Kind
get_posture_summary	Headline counts: total, managed, unmanaged and drifted resources, plus misconfigurations	Read
get_resource_breakdown	Resource counts grouped by type, region, account, cloud or status	Read
get_top_misconfigurations	The resources with the most misconfigurations	Read

Search and discovery

Find workflows, stacks, runs, policies, templates, connectors and more. See Search and discovery tools.

Tool	What it does	Kind
search_workflows	Find workflows by name, status, type, template, connector, drift, policy and more	Read
search_stacks	Find stacks by name, status, template, drift or policy	Read
search_workflow_runs	Find runs by status, workflow, trigger or time	Read
search_policies	Find policies by name or by the template they use	Read
search_orchestrator_resources	Search any resource type, including runners, roles, secrets and connectors	Read
search_marketplace_templates	Find templates in your organization's library	Read
get_orchestrator_resource_by_id	Full details of one workflow, stack, run or other resource	Read
get_marketplace_template_by_id	Full details of one template or template revision	Read
cross_reference	Answer "what uses this?" for a template, connector or runner group	Read
aggregate_orchestrator_resources	Count resources grouped by status, type, environment and more	Read

Workflow runs and results

Run workflows, act on runs in progress, and get logs, outputs and artifacts. See Workflow run tools.

Tool	What it does	Kind
get_workflow_run	Read one run's current status and details directly from the platform	Read
get_workflow_run_logs	A run's console output, optionally only the errors	Read
get_workflow_outputs	The outputs a workflow or stack published, with sensitive values redacted	Read
list_workflow_artifacts	The files a workflow's runs produced	Read
get_workflow_artifact	A temporary download link for one artifact file	Read
create_workflow_run	Start a run: plan, apply, refresh, destroy and more	Write
control_workflow_run	Approve, reject, revoke an approval on, or cancel a run	Write

Managing workflows

Create, change and remove workflows. See Workflow management tools.

Tool	What it does	Kind
create_workflow	Create a workflow from a library template or a git repository	Write
update_workflow	Change a workflow's settings, or upgrade it to a new template revision	Write
delete_workflow	Delete a workflow's records (doesn't destroy infrastructure)	Write
upload_workflow_tfstate_url	Get a link to replace a workflow's Terraform state with a file you provide	Write

Managing stacks

Create, configure and run bundles of workflows, and follow their progress. See Stack tools.

Tool	What it does	Kind
create_stack	Create a stack and its workflows from a pinned stack template	Write
update_stack	Change a stack's settings, schedules or action definitions	Write
create_stack_run	Start one of a stack's actions	Write
get_stack_run	Read a stack run's progress and its child workflow runs	Read
delete_stack	Delete a stack's records, optionally including its workflows	Write

Templates

Manage the IaC templates in your library and see your marketplace subscriptions. See Template tools.

Tool	What it does	Kind
get_subscription	The marketplace templates your organization has subscribed to	Read
create_workflow_template	Register a git repository as an IaC template	Write
update_workflow_template	Edit a template, or configure or publish one of its revisions	Write
create_workflow_template_revision	Add a new version to an existing template	Write
delete_workflow_template	Delete one template revision, or a template record with no revisions	Write

Policies

Create and change governance policies. See Policy tools.

Tool	What it does	Kind
create_policy	Create a policy with rules, verdicts and a scope	Write
update_policy	Change a policy's rules, scope or settings	Write

Secrets and connectors

Manage vault secrets and check connector credentials. See Secret and connector tools.

Tool	What it does	Kind
create_secret	Store a new secret in your organization's vault	Write
update_secret	Rotate an existing secret's value	Write
delete_secret	Permanently delete a secret, after an impact check	Write
authenticate_integration	Check that a connector's stored credentials still work	Live check

Presenting answers

Tool	What it does	Kind
render_blocks	Present an answer as UI blocks in a compatible client such as Kyro	Read

render_blocks

Presents information the assistant has already gathered as StackGuardian UI blocks, such as a table or chart, in a client that supports them. It doesn't query data or change platform resources.

The assistant uses this tool only when its client supplies the rendering instructions. In clients without those instructions, it answers in plain text or Markdown instead.

What the tools can't do

• Complete historical run facts. There is no dedicated tool for fetching a historical run-facts document. A resource's by-id response can include stored policy outcomes and cost fields where populated, but this isn't a complete facts lookup for any chosen run. Use the dashboard for run-specific cost and policy details.
• Other organizations. Tools act only in the organization you connected with and never take an organization from the assistant.
• Reading stored vault-secret values or artifact contents. Vault-secret values can't be read back, and artifact tools return links instead of file contents. Sensitive values can still appear in logs or plain-text configurations, see How the tools work.