Test the application.
Review the evidence.
Teach your coding agent your application’s rules, have it run repeatable tests, and review the changes in your pull request.
Run tests across
your application.
A flow can combine browser actions with database checks and commands from your own plugins. Declare the setup it depends on, inspect the execution plan and compare the results.
Explore the configuration ┌──────────────┐
│ seed-data │
└──────┬───────┘
│
┌─────────────┴─────────────┐
│ │
┌──────▼───────┐ ┌──────▼───────┐
│ invite-member│ │ edit-profile │
└──────┬───────┘ └──────────────┘
│
┌──────▼───────┐
│ verify-role │
└──────┬───────┘
▼
┌──────────────┐
│ accept-invite│
└──────────────┘ Application knowledge
Teach your agent your application’s terminology and business rules, with references to the flows that test them.
A change to team permissions has a place to start: the invitation flow, its prerequisite data, and the role checks that depend on it.
Explore application knowledgeknowledge / team-access.md
│
▼
┌──────────────┐
│ seed-data │
└──────┬───────┘
▼
┌──────────────┐
│ invite-member│ ◄── application rule
└──────┬───────┘
▼
┌──────────────┐
│ verify-role │
└──────────────┘DAG execution
A Terraform-like loop for your agent to run application tests: declare dependencies, inspect the work, and execute repeatable flows.
Prepare data once, run independent journeys in parallel, and keep dependent checks behind their prerequisites. Reuse eligible completed evidence when inputs and execution state still match.
Explore dag execution ┌───────────┐
│ seed-data │
└─────┬─────┘
┌───────┴────────┐
▼ ▼
┌─────────────┐ ┌─────────────┐
│invite-member│ │edit-profile │
└──────┬──────┘ └─────────────┘
▼
┌─────────────┐
│ verify-role │
└─────────────┘Plugins
Test business journeys across your browser, services, databases and background jobs. Give your agent commands to run them and comparisons to show what changed.
Turn “a refund restores the customer’s balance” into a repeatable flow. Your agent can run the business journey and bring changes to its results into PR review.
Explore plugins your business rules
│
▼
application commands
├── place an order
├── issue a refund
└── capture the balance
│
▼
compare the results
├── amount refunded
├── stock restored
└── customer balance
│
▼
PR reviewBrowser runs
Open pages, follow user journeys, assert behavior, and keep screenshots at the points that matter.
The reviewer can see whether the invitation worked and inspect the screen where it failed.
Explore browser runs open /settings/team
│
▼
click "Invite member"
│
▼
fill email / choose role
│
┌────┴────┐
▼ ▼
assert snapshot
└────┬────┘
▼
retained resultBrowser deltas
Give your agent snapshot comparisons to judge which application changes belong in the scope of your PR.
A PR meant to update invitation copy also removes the role selector. Your agent flags the unexpected removal, links the snapshots, and explains why it appears outside the intended scope.
Explore browser deltas ┌────────────┐ ┌────────────┐
│ BEFORE │ │ AFTER │
│ email │ │ email │
│ role [ v ] │ │ │
│ [ invite ] │ │ [ invite ] │
└─────┬──────┘ └──────┬─────┘
└─────────┬────────┘
▼
role selector removed
│
▼
screenshots + PR noteEvidence deltas
Compare changes to screenshots, database results, command output and application knowledge in the same review.
An interface change and the database state behind it belong in the same review. Your agent can explain the relationship and link back to the originals.
Explore evidence deltas BEFORE AFTER
────── ─────
screen ────────── screen ┐
rows ────────── rows ├─► deltas
output ────────── output │ │
knowledge ─────── knowledge┘ ▼
PR review
│
inspect originals
│
human decisionPR review & comments
Help your reviewer understand what ran, what changed, and what still needs attention.
The review starts with changed evidence and unresolved behavior. Reviewers can open the originals and make the merge decision in GitHub.
Explore pr review & comments PR head
│
▼
plan → execute → retain evidence
│
┌────────┴────────┐
▼ ▼
progress comment visual report
│ │
└────────┬────────┘
▼
your human review