What it does
This is the main design-time workspace for creating workflows. Instead of asking for code in a vacuum, GTMship treats generation like an engineering loop: inspect active connections, ground the relevant APIs, generate a draft, then keep iterating through validation, preview, build, and deploy.
What improves the first draft
Clear outcome and trigger
The best prompts explain what starts the workflow, what result should happen, and what success looks like for the operator or downstream team.
Known systems and objects
Naming the providers, records, and actions involved helps the agent ground the right APIs instead of making broad guesses.
Useful sample inputs
Example payloads, field names, or object IDs make branching logic and data mapping much more concrete during generation.
Safety expectations
Telling the agent which steps should stay read-only in preview or require approval helps it shape checkpointed writes from the start.
How it works
- You describe the workflow outcome in plain language.
- The agent checks what connections are already available and what external systems the draft will need.
- It reads integration references, docs, or OpenAPI details so it can ground endpoint usage before drafting code.
- It generates or revises a workflow artifact, keeping code, flow representation, and supporting notes together.
- You continue the loop with validation, preview, build, deploy, or further revision instead of restarting from scratch.
Prompt
-> check active connections
-> ground API details
-> generate workflow draft
-> validate + preview
-> revise or deploy
What the draft includes
TypeScript workflow code
The generated draft is real workflow source shaped around GTMship's SDK and runtime helpers.
Flow representation
A diagram view helps you inspect triggers, branches, and downstream writes without reading every line of code first.
Validation context
The draft stays tied to validation results so compile issues or missing checkpoints can be repaired in the same workspace.
Preview and deploy context
The same artifact moves forward into preview, build, and deployment planning instead of being copied between tools.
How revision stays grounded
- Missing integrations stay explicit: if a needed provider is not connected, the next step is to add or configure it instead of pretending the dependency is already solved.
- Validation feedback stays attached: compile issues, schema problems, and checkpoint gaps can be fixed against the same workflow artifact.
- Preview feedback is concrete: reads, branches, and paused writes give the agent something specific to revise instead of relying on vague prompt tweaks.
- Deployment planning reuses the same draft: the workflow that was grounded and previewed is the one that moves into packaging and binding resolution.
What you still control
- The workflow prompt and any revision instructions.
- Which providers are connected and available to the draft.
- Whether preview writes are approved or blocked.
- Whether the final artifact is shipped to cloud infrastructure.