What's new in Arty — April 2026

A summary of everything shipped in April 2026 across Studio, the Agent Library, and the API. Smaller improvements, fixed edges, and a few things that have been on the list longer than we would like to admit.

Studio

Workflow run history is now searchable. You can filter past runs by agent, status, date range, or a keyword from the input or output. Previously you could only scroll. If you run high-volume agents, this makes a real difference.

Conditional logic in workflows now supports nested rules. Before this release, you could define one condition per workflow step. You can now nest up to three conditions with AND/OR operators. The UI is a simplified rule builder — no code required.

Agent config drafts are auto-saved. If you close Studio mid-configuration, your work is now saved as a draft automatically. It will be waiting for you when you come back.

Agent Library

Two new agents shipped this month.

Contract review agent — surfaces non-standard clauses, flags missing sections, and produces a plain-language summary of a contract document. Works on PDFs and Docs. Currently in beta for Scale and Enterprise customers; available to Growth in May.

Employee onboarding agent — guides new hires through a structured onboarding workflow, sends task reminders, collects confirmations, and posts a completion summary to your HRIS. Integrates with Workday, BambooHR, and Rippling out of the box.

HubSpot integration now supports bidirectional sync. Previously, agents could read from HubSpot and write enrichment data back. You can now also trigger agent runs directly from HubSpot workflow actions. Set it up from the Integrations page in Studio.

API

Streaming responses are now available on the /v1/runs endpoint. For agents that return long-form text output — drafts, summaries, briefs — you can now stream the response token by token instead of waiting for the full completion. Pass "stream": true in your request body.

{
  "agent_id": "agt_01hx4z9k2m",
  "inputs": [{ "domain": "acme.io" }],
  "stream": true
}

{
  "agent_id": "agt_01hx4z9k2m",
  "inputs": [{ "domain": "acme.io" }],
  "stream": true
}

{
  "agent_id": "agt_01hx4z9k2m",
  "inputs": [{ "domain": "acme.io" }],
  "stream": true
}

Webhook retry behaviour has been updated. Failed webhook deliveries now retry with exponential backoff: 10s, 30s, 2m, 10m, 1h. Previously retries were fixed at 60-second intervals. The delivery log in your dashboard shows the full retry history for each event.

A new /v1/agents/:id/stats endpoint is available. Returns run count, average latency, success rate, and error breakdown for any agent over a configurable time window. Useful for building internal monitoring dashboards.

Fixes

• Fixed an issue where agent run outputs were occasionally truncated when the response exceeded 8,000 tokens. Limit is now 32,000 tokens.

• Fixed a display bug in Studio where the workflow canvas would not render correctly on screens narrower than 1,280px.

• Fixed incorrect time zone handling in scheduled agent runs for users in UTC+5:30 and UTC+5:45.

• Resolved a race condition in the HubSpot integration that caused duplicate CRM writes under high concurrency.

The full changelog with every commit is on the Changelog page. Next month: expanded competitive monitoring capabilities and the first public release of the Arty CLI.