Tutorials

Mastra Claims Demo

Run the bundled Mastra claims example and use it as a complete capture and replay playground.

The repository includes a claims-agent example under resources/examples/mastra-claims-agent. It is the best local demo because it exercises domain data, tools, capture, and an investigation script.

1. Start isplay

npx isplay start
export ISPLAY_API_URL=http://127.0.0.1:7373
npx isplay projects create --name "Mastra Claims Demo"
export ISPLAY_PROJECT_ID="<project id>"

2. Run The Demo

From the repository root:

npm run demo:mastra-claims

The demo records a claims workflow run and prints IDs needed for inspection.

3. Inspect The Captured Run

npx isplay runs list --project "$ISPLAY_PROJECT_ID"
npx isplay discover run <runId>
npx isplay runs context <runId>
npx isplay runs checkpoints <runId>

Look for:

  • Mastra step events.
  • Model call projections.
  • Tool proposal and execution records.
  • Domain state checkpoints.
  • Context inventory that explains the decision.

4. Try One Hypothesis

Choose one prompt clause, retrieval item, or state field from the context inventory. Then create one branch and replay it:

npx isplay branch create \
  --from <runId> \
  --checkpoint <checkpointId> \
  --project "$ISPLAY_PROJECT_ID" \
  --name "demo-hypothesis"

npx isplay branch intervene <branchId> \
  --project "$ISPLAY_PROJECT_ID" \
  --kind prompt_clause_mask \
  --target-context <contextItemId> \
  --description "Mask one policy clause"

npx isplay replay <runId> \
  --project "$ISPLAY_PROJECT_ID" \
  --branch <branchId> \
  --model-policy recorded-only \
  --tool-policy pause-for-fixture

5. Use It As A Template

The example files show the pieces most real apps need:

FileUse
src/agent.tsAgent workflow and model/tool boundary.
src/tools.tsTool implementation and tool capture concerns.
src/data.tsDomain fixture data.
src/isplay-lab.tsisplay setup and investigation helpers.
src/run.tsDemo entrypoint.

When porting the pattern, keep checkpoint names and context paths stable. That is what makes later replay and experiment specs readable.

On this page