Three ways teams use Arga.

Training environments for RL labs, sandboxes for enterprise agents, and validation for code changes.

acme / workspaceAK

Refund escalation

preview/refund-policy

Running

Plan

3 / 4

Open refund request

Read customer history

3Apply escalation policy

4Verify side effects

Live trace

12:04:21.008 browser.open /refunds/1842

12:04:22.413 gmail.users.messages.get 200

12:04:23.071 slack.chat.postMessage intercepted

12:04:23.716 policy.evaluate escalation=true

12:04:24.118 verifier.check_side_effects running

RL training environments

Arga gives models multi-app workflows where they can practice real tool use, acquire skills, and improve before touching production.

acme / workspaceAK

Twin Runs

Spin up short-lived or permanent service twins without deploying app code.

Service twins

API-only

API + UI

Session type

Short-lived sessionPermanent

Expires automatically after 60 minutes.

Describe the twin state (optional)

Describe the state your agent should encounter.

Provision twins

Enterprise agent sandboxes

Run agents against realistic twins of services like Slack, Stripe, GitHub, and Jira, with ground truth of good tool calls.

acme / workspaceAK

PR Test Runs

Set up repository tests, review recent PR runs, and track the latest test status from one place.

PR test setup

Run tests on pull requests and publish the result as the Arga Validation check.

Repository

argalabs/support-agentConnected

Trigger mode

Pull requests

Post a bot comment

Comment when a PR test run finishes.

Custom instructions (optional)

Run the saved refund escalation tests and verify Slack and Gmail side effects.

Save validation settings

Code-change validation

Arga runs a deterministic browser runner that clicks through the changed flows and validates functionality against integrations.

acme / workspaceAK

Refund escalation

preview/refund-policy

Running

Plan

3 / 4

Open refund request

Read customer history

3Apply escalation policy

4Verify side effects

Live trace

12:04:21.008 browser.open /refunds/1842

12:04:22.413 gmail.users.messages.get 200

12:04:23.071 slack.chat.postMessage intercepted

12:04:23.716 policy.evaluate escalation=true

12:04:24.118 verifier.check_side_effects running

RL training environments

RL Training Environments

Arga gives models multi-app workflows where they can practice real tool use, acquire skills, and improve before touching production.

See how researchers use Arga

acme / workspaceAK

Twin Runs

Spin up short-lived or permanent service twins without deploying app code.

Service twins

None selected

API-only

Salesforce

Jira

Twilio

API + UI

GitHub

Slack

Stripe

Gmail

Google Calendar

Session type

Short-lived sessionPermanent

Expires automatically after 60 minutes.

Describe the twin state (optional)

Describe the state your agent should encounter.

Provision twins

Enterprise agent sandboxes

Run agents against realistic twins of services like Slack, Stripe, GitHub, and Jira, with ground truth of good tool calls.

See how agents use Arga

acme / workspaceAK

PR Test Runs

Set up repository tests, review recent PR runs, and track the latest test status from one place.

PR test setup

Run tests on pull requests and publish the result as the Arga Validation check.

Repository

argalabs/support-agentConnected

Trigger mode

Pull requests

Post a bot comment

Comment when a PR test run finishes.

Custom instructions (optional)

Run the saved refund escalation tests and verify Slack and Gmail side effects.

Save validation settings

Code-change validation

Arga runs a deterministic browser runner that clicks through the changed flows and validates functionality against integrations.

See how engineers use Arga

Make your agents more capablein the real world.

Spin up isolated environments with API twins, full traces, and production-data for apps and agents that need to act.

Spin up a twin