AWS-shaped resource types
140 resource types, each named the way CloudFormation names it (AWS::EC2::VPC, AWS::S3::Bucket, AWS::RDS::DBInstance) with the same property names. If you know the schema, you know Foundry's.
Foundry provisions AWS resources with the SDK, plans every change against what is actually live, and keeps an inventory, a compliance score, advisor findings, cost, and IP space for every account in the organization. Renkara runs its own AWS accounts on it. It is in a closed preview: ask for early access by email.


Renkara runs its own AWS accounts on Foundry today. If you want it for yours, join the waitlist by email and we will write to you when it opens. There is no price yet and no launch date.
Everything below is in the code today, counted from it on 10/29/2026.
140 resource types, each named the way CloudFormation names it (AWS::EC2::VPC, AWS::S3::Bucket, AWS::RDS::DBInstance) with the same property names. If you know the schema, you know Foundry's.
Every type is a Python class that creates, reads, updates, and deletes through the AWS SDK. No Terraform, no CloudFormation templates, no CDK.
A plan is the typed diff of what you declared against what is live. Apply runs the plan, destroy tears a stack down, and a converged stack plans to nothing.
Stack changes are saved as immutable versions you can list, diff, and roll back to. A release groups changes across stacks through plan, approval, apply, and rollback.
Import live resources with their physical IDs attached, so the next plan on the imported stack is empty and nothing is rebuilt.
132 discoverers describe the organization's resources across every account Foundry knows, in three regions by default, eight account and region pairs at a time. Scans start on demand.
979 rules over 28 resource types, evaluated against the stored inventory, with a finding per failing resource. Coverage is partial: about half the rules read properties no discoverer collects yet.
Trusted Advisor style checks written in house: 46 security, 24 cost, 14 service limits, 13 fault tolerance, and 5 performance, each with severity, affected resources, and an estimated saving where one applies.
Cost Explorer first, the Cost and Usage Report when it is provisioned. Actual and estimated spend are reported separately, and a total is never shown without the coverage behind it.
Thirty, 60, and 90 day spend forecasts from a linear trend, and a flag on any day more than two standard deviations out. Findings on managed resources become governed rightsizing patches.
Free and used IPs per subnet, a ledger of every VPC and subnet CIDR, overlap checks, suggested blocks from a supernet budget, a peering and transit gateway map, and an exhaustion forecast per VPC.
Define the tag keys and rules, find untagged and expiring resources, chargeback by taxonomy, and see which resources violate which policy.
Before an apply, Foundry scores every resource that depends on what is changing, and blocks a high-risk apply until an operator dismisses the report. The CIDR gate blocks overlapping creates. Both are configuration switches.
A policy binds a finding to a corrective stack patch. A curated safe list applies automatically; everything else opens an approval card in Docket.
Cluster an unmanaged account's inventory into dependency-coherent stacks, import each with its physical IDs, and report managed coverage.
Stack versions, deployments, and CloudTrail events fused into one timeline per resource, with the property state reconstructed at any past moment and a single-resource rollback.
A shadow of a stack in a disaster recovery account and region, with zone, image, CIDR, and ARN values rewritten, a readiness score, and a failover call.
Describe a stack and get one back that matches the resource schemas, wired and cycle-checked, with a plan ready. Types outside the registry are rejected, not guessed.
Real captures of the running tool, with fictional StrataForge Robotics data. They follow your theme.
















Declare a stack as a flat list of resources with CloudFormation-shaped types and properties. There is no template language and no module hierarchy.
Foundry resolves dependencies and diffs the declaration against live state. Every create, update, replace, and destroy is listed in order before anything happens.
boto3 executes the plan through the governance and safety gates, and the deployment is recorded with its plan and per-resource outcome.
Scans keep the inventory current. Compliance, advisor, cost, and IP space recompute from it, and remediation policies turn findings into governed patches.
This website is scanned with axe-core in light and dark at phone and desktop widths and has no serious or critical findings. The Foundry console started at 94 serious or critical findings in the same scan of 17 screens; after the repairs of 10/29/2026 the scan finds none, in light and dark, at phone and desktop widths, and we list what is still open below. We do not claim a WCAG conformance level.
The console has a skip link to the main content, a main landmark, and labeled navigation regions. Every scrolling region (tables, JSON blocks, lists) is a named region a keyboard can reach, and the inventory table's header checkbox is its own control. A keyboard-only walkthrough of every flow has not been recorded.
Landmarks are labeled, every icon button has a name, and axe-core finds no naming or label violation on any scanned screen. Nothing has been tested with a screen reader.
Text contrast passes in light and dark. The console has real light, dark, and system themes on design-system tokens, a token-level test checks every text and tint pair at 4.5 to 1 in both, and the axe scan finds no color-contrast violation on any of the 17 screens. It was checked by tools, not by a person with low vision.
The Renkara design system carries a reduced-motion rule that the console inherits. Foundry's own tests do not check it.
Resizing text to 200 percent has not been tested.
At 375 by 667 pixels none of the 17 scanned console screens scrolls sideways, in either theme. The side navigation is a drawer below 768 pixels, and tables scroll inside labeled regions. It was checked in a desktop browser at that width, not on a phone.
| Backend |
|
|---|---|
| Console |
|
| AWS integration |
|
| Rules and checks |
|
| Agent interface |
|
Remediation opens an approval card in Docket for anything outside the safe list.
When monitoring is switched on, every apply and destroy reconciles the Vigil monitors for the resources it changed.
Foundry ships its logs and traces to Chronicle.
Narrative sites are described to Foundry through a service-token handshake.
Everything the console does is also a tool. An agent can plan, release, scan, query, and remediate through the same role checks and gates a person meets. A selection of the tools:
foundry_list_stacksfoundry_get_stackfoundry_create_stackfoundry_plan_deploymentfoundry_apply_deploymentfoundry_destroy_stackfoundry_detect_driftfoundry_list_stack_versionsfoundry_rollback_stackfoundry_add_resourcefoundry_update_resourcefoundry_validate_resourcefoundry_list_resource_typesfoundry_create_releasefoundry_plan_releasefoundry_approve_releasefoundry_apply_releasefoundry_query_inventoryfoundry_inventory_summaryfoundry_scan_fleetfoundry_get_scanfoundry_untagged_inventoryfoundry_resource_timelinefoundry_list_compliance_packsfoundry_run_compliance_packfoundry_list_compliance_findingsfoundry_run_advisorfoundry_list_advisor_findingsfoundry_advisor_savings_summaryfoundry_cost_summaryfoundry_cost_forecastfoundry_cost_anomaliesfoundry_cost_by_taxonomyfoundry_stack_costfoundry_chargeback_reportfoundry_cidr_check_conflictfoundry_cidr_suggestfoundry_cidr_topologyfoundry_cidr_exhaustion_forecastfoundry_vpc_ip_summaryfoundry_taxonomy_listfoundry_governance_coveragefoundry_governance_enforcement_violationsfoundry_create_remediation_policyfoundry_evaluate_remediation_policiesfoundry_approve_remediationfoundry_synthesize_stackfoundry_adoption_startfoundry_reconstruct_statefoundry_create_stack_mirrorfoundry_stack_mirror_failoverfoundry_simulate_blast_radiusfoundry_access_graph_run_analysisFoundry is currently in a closed engineering preview. The Solo and Team tiers open to everyone when it leaves the preview; join the waitlist and we'll email you the day they do. Enterprise onboarding is currently active on a rolling basis.
Email contact@renkara.com with the subject "Foundry early access" and a line about what you manage. There is no form on this site.
Not yet. Nothing here states one, and no checkout exists.
No. Every screenshot is a capture of the running app against a fictional StrataForge Robotics AWS organization, with no real account, resource, or cost data.
For the accounts it manages, yes: it provisions through the AWS SDK and does not use Terraform, CloudFormation templates, or CDK. It can import resources that already exist.
Partial. Foundry ships 129 packs and 979 rules, but about half the rules read properties no discoverer collects yet, so they cannot pass.
Write to Renkara Media Group and a person will answer.