MCP Server

MicroStax exposes a first-class MCP server for agent workflows. The current implementation registers:

• 40 tools
• 5 resources
• 3 built-in prompts

This page documents the active server surface created by createMcpServer().

Connection model

The MCP server is a task-oriented interface over the same control plane used by the CLI and dashboard.

Typical configuration:

• API URL from MCP server config
• bearer token for authenticated actions
• optional organization scope for org-specific tasks

Tool Catalog

Environment tools

Tool	Inputs	Purpose
env_list	status?	List environments, optionally filtered by status.
env_get	id	Fetch environment details.
env_create	blueprint, name?	Create a new environment from Blueprint content.
env_delete	id	Delete an environment.
env_stop	id	Stop an environment.
env_resume	id	Resume an environment.
env_status	id	Fetch live service status.
env_update	id, name?, autoShutdownEnabled?, autoShutdownDelayMinutes?	Update mutable environment settings.
env_scale	id, service, replicas	Scale a service.
env_replay	id, snapshotId?	Replay an environment execution graph.
env_get_diffs	id	Fetch behavioral diffs for an environment.

Blueprint tools

Tool	Inputs	Purpose
blueprint_validate	blueprint	Validate Blueprint content before deployment.
blueprint_discover	path?	Discover a Blueprint from source code.

Logs, topology, metrics, and drift tools

Tool	Inputs	Purpose
env_logs	id, service?, lines?	Fetch recent logs.
env_topology	id	Return topology graph data.
env_metrics	id	Return resource metrics.
env_deviations	id	Return configuration drift data.
env_cost_analysis	id	Return sparse cost analysis and savings.

Diagnosis and remediation tools

Tool	Inputs	Purpose
env_diagnose	id	Run AI diagnosis.
env_diagnose_apply	id, suggestionId	Apply a diagnosis suggestion.

Snapshot and data tools

Tool	Inputs	Purpose
snapshot_create	id, service?	Create a snapshot.
snapshot_list	id	List snapshots for an environment.
env_seed	id, service, script, type	Run an inline seed script against an environment.
seed_search	query	Search the seed registry.
seed_run	name, envId, version?	Run a published seed.
seed_publish	name, version, content, serviceType	Publish a seed artifact.

Baseline and overlay tools

Tool	Inputs	Purpose
baseline_promote	id, force?	Promote an environment to baseline.
overlay_create	baselineId, name, blueprint	Create an overlay from a baseline.
overlay_delete	overlayId	Delete an overlay.

Sharing, template, and stub tools

Tool	Inputs	Purpose
env_share	id, expiresInHours?	Create a shareable link.
template_list	none	List templates.
template_get	id	Fetch one template.
stub_generate	spec	Generate AI service stubs from an API spec.

Health, billing, and organization tools

Tool	Inputs	Purpose
health_check	none	Quick API health check.
health_details	none	Detailed system health response.
billing_get	none	Billing and plan summary.
org_get	id?	Get organization details.
org_analytics	none	Organization analytics summary.
org_compliance_report	orgId	Compliance report.
governance_logs_get	none	Governance audit log summary.

Resources

The active server registers the following MCP resources:

Resource	URI	Purpose
environment_list	microstax://environments	Current environment list as JSON.
environment_detail	microstax://environments/{id}	One environment record as JSON.
environment_logs	microstax://environments/{id}/logs	Environment logs as plain text.
blueprint_schema	microstax://schema/blueprint	Blueprint schema resource.
template_list	microstax://templates	Template inventory as JSON.

Built-In Prompts

The active server exposes these built-in prompts:

Prompt	Inputs	Intended flow
create-environment	description	Generate Blueprint, validate, create environment.
diagnose-failures	environment_id	Fetch details, inspect logs/topology, run diagnosis.
blueprint-review	blueprint	Review Blueprint content before deployment.

Typical agent workflow

Create and verify an environment

1. read microstax://schema/blueprint
2. generate Blueprint text
3. call blueprint_validate
4. call env_create
5. poll with env_get or env_status

Diagnose a broken environment

1. call env_get
2. call env_status
3. call env_logs
4. call env_topology
5. call env_diagnose
6. optionally call env_diagnose_apply

Data and validation workflows

1. call snapshot_create
2. call seed_search or env_seed
3. call env_get_diffs
4. call env_cost_analysis

Practical guidance

• use MCP when the caller wants task semantics rather than low-level REST orchestration
• use resources for read-mostly context such as schema, template inventory, and environment state
• use prompts to standardize agent behavior for environment creation and diagnosis
• prefer the REST API or CLI when a workflow depends on exact transport-level control or large-scale scripting