Book a demo

Tell us about your use case. We will walk through composing a workflow, deploying to AWS Bedrock AgentCore, and invoking it from your product.

Workflows

Create, version, deploy, and run AI workflows in Floprints. Lifecycle from composing through ready, with immutable version history.

Blueprint lifecycle

A blueprint (workflow) moves through these statuses:

StatusMeaning
composingActive composer session; plan still being designed
draftPlan saved but not yet deployed
deployingHarnesses being created or reconciled
readyDeployed version available for runs
failedDeploy or compose failed

Versions

Every compose or redeploy creates an immutable workflow_version containing:

  • Full WorkflowPlan
  • Model used by the composer
  • Tools snapshot
  • System-prompt hash
  • Edit summary
  • Complete composer transcript

Version lifecycle: draft → deploying → deployed → superseded (or failed).

A blueprint tracks currentVersionId and deployedVersionId independently — enabling rollback and side-by-side history.

Console surfaces

PagePurpose
Workflows listAll blueprints with status and last activity
ComposeConversational design session
DetailPlan view, version history, deploy controls
RunExecute and chat with a deployed workflow
Run detailStep trace, messages, token usage, human requests

Runs

Each execution creates a run with:

  • Status: pending, running, completed, failed, timed_out
  • Phase: adds waiting_human when paused for input
  • Per-agent steps with input/output and timing
  • Full message trace (roles, content, stop reason, token usage, latency, reasoning)
  • Human requests when the workflow asks a question

Orchestrator runs can be continued via follow-up turns that append to the persistent session.

Background poller

A 5-second background worker resumes runs left in running or waiting_human: applies human answers, performs crash-recovery re-invocation, and guards against double-invocation.