Step Functions is one of the best services AWS has, and one of the hardest to operate at scale. The console shows one state machine in one Region at a time. Finding the executions that failed overnight means opening each workflow in turn. A history of thousands of events is read a page at a time. Redrive is one execution at a time, and nothing records who started or stopped what.
StateCraft is a console for Step Functions that fixes the day-to-day parts. It shows every state machine across your accounts and Regions, explains each execution state by state, redrives failures in bulk, and comes with a workflow designer you can drive with the mouse, the keyboard or your voice. It runs entirely in your own AWS account.
Runs in your account
StateCraft is one serverless CloudFormation stack: Amplify Hosting for the web app, Amazon Cognito for sign-in, an HTTP API with Go Lambda functions, and DynamoDB. It costs close to nothing when nobody is using it.
- Nothing goes to us. Definitions, inputs, outputs, histories and the audit log stay in your account. Cloud Certainty runs no service in the data path, and the app sends no telemetry.
- Your identity provider. Sign in through IAM Identity Center, any SAML or OIDC provider, or Cognito users with MFA.
- Other accounts. Connect them with a small role in each, rolled out with a StackSet if you like. The role sets the ceiling: Read, Operate or Full.
The fleet, across accounts and Regions
The home page lists every state machine you may see, in every connected account and Region, with executions started, failures, success rate and p95 duration from CloudWatch, a sparkline of the period, and a "failing now" flag. Filter by account, Region, type or name, keep favourites, and share the view as a link.
Executions you can actually read
Open an execution and you get:
- A live graph with the status of every state, the path the execution took, durations, retries and Map iteration counts. Select a state for its attempts, backoff, input, output and what changed between them.
- A timeline with every state as a bar, each retry as its own bar, and lanes for Map iterations and Parallel branches.
- The full event history, loaded a thousand events at a time up to the 25,000-event limit, with filters, search and a compact view, and a warning when an execution gets close to the limit.
- A Map Run explorer for Distributed Map: progress, child executions, failures grouped by error and cause, and live changes to concurrency and failure tolerance.
- Compare two executions of the same workflow: their paths, the time each state took, and their inputs and outputs, side by side.
Each execution also shows an estimated cost from its state transitions, why it can or cannot be redriven, and, when it was started through StateCraft, who started it.
Search and bulk actions
Search failed, timed out or aborted executions across up to 200 state machines at once, grouped into failure clusters. Select them and stop, redrive or re-run them as one bulk job. StateCraft runs bulk jobs on Step Functions itself: a Distributed Map in your account works through them, under the identity of the person who asked.
A designer you can talk to
The designer draws your workflow as you build it, with a two-way code view, forms for every state type in JSONata or JSONPath, retry and catch editors, and checks as you type, plus AWS's own validation. Run a single state with TestState, try JSONata expressions on sample data, then deploy with a version and an alias.
The palette holds every AWS SDK action Step Functions can call: 411 services and about 18,800 actions, grouped and searchable. And you can say what you want: "add a Lambda task called Charge card after Validate order", "if amount is greater than 100 go to Manual review otherwise go to Approve", "retry Charge card 3 times with backoff". Speech goes to Amazon Transcribe in your own account, and every command shows what it changed and can be undone.
Grants that AWS enforces
Admins grant view, data, run and deploy per user or group on state machine name patterns such as orders-*. Without data, StateCraft asks Step Functions for metadata only, so inputs and outputs stay hidden. Every request runs with a short-lived session narrowed by a session policy to the caller's grants, so AWS refuses anything outside them even if StateCraft had a bug. CloudTrail names the person: every session carries their email as its source identity.
Early access
StateCraft is in early access ahead of its AWS Marketplace launch. It will be sold like our other products: priced per deployment, not per user, and licensed through AWS License Manager in your account.
Read more on the StateCraft page (opens in a new tab) and in the docs (opens in a new tab), or email [email protected] to join the early access programme.