Changelog
Follow up on the latest improvements and updates.
RSS
We’ve updated Gap Finder to make it easier to identify and fill knowledge gaps in your AI Agent.
What’s new:
- Gap Finder now highlights New This WeekandTop Knowledge Gapsso you can quickly prioritize the latest unanswered questions and the gaps with the biggest impact.
- A clearer card-based layout makes each gap easier to scan, rank, and open for editing.
- Use View allto open the All Gaps page, where you can search, filter, and page through the full list of gaps.
These updates are designed to help teams prioritize the right gaps faster and improve resolution rates with less manual digging.
new
Chat History
We've added a new widget feature - chat history ⏳.
From now on, we will store last 5 chats made by the user. They will be able to access them by clicking three dots in the top-right section and then selecting
Chat History
. Chat will be archived automatically after 2 hours of inactivity and only when widget initializes (for instance, when user refreshes the page).You can now send AI agent escalation submissions to any HTTPS endpoint with a custom webhook.
How to use it:
- Go to Configure -> Escalation
- Choose Webhookas the contact method
- Add your endpoint URL and configure the request
- Choose which form fields and contact data to send in the request body
- Set a fallback method in case the webhook request fails
Why it's useful:
- Connect Helply to your own internal tools, automations, or custom support workflows
- Send customers’ form responses and key contact details exactly where your team needs them
- Support teams that don’t use one of Helply’s helpdesk integrations
- Keep escalations from getting lost by setting a fallback destination if the webhook fails
Contact Mocking lets you simulate how your AI agent behaves for different types of customers directly from the Playground without modifying any real contact data. Fill in custom field values (like plan type, company name, or subscription tier) and see how your agent responds in real time, using the exact same pipeline as production.
How to use it
- Open your agent's Playground page
- Fill in the contact fields in the Contact Mocking form (fields come from your existing Custom Fields schema)
- Start chatting the agent will respond as if it's talking to a real contact with those attributes
- Save commonly used field combinations as Presets for quick reuse
Test your agent's behavior across different customer scenarios before they happen in production. 🚀
Take control of your monthly resolution costs with the new Resolution Spend Cap. Set a limit on how much you're willing to spend on resolution overages each billing cycle, so there are no surprises on your invoice.
How to use it
- Go to Settings → Billingand enter an amount as your monthly spending cap.
- To only receive a warning, leave the "Widget Control" toggle off. You'll get an email alert when you reach 80% of your cap, but your chatbot will continue responding normally.
- To pause AI responses when the limit is reached, simply toggle the "Widget Control" button on. Your chatbot will stop generating AI responses once you hit the cap, and conversations will be routed to your team instead.
- If your chatbot gets paused, increase the cap or set it to $0to unpause it. AI responses resume immediately.
- To opt out of spend alert emails, toggle off the email preference in your Billing settings. The email alert setting applies to each user individually; disabling it for yourself won't impact other account owners.
Why it's useful
If your account uses resolution-based billing, costs can vary month to month depending on chatbot activity. The spend cap gives you a predictable ceiling so you always know your maximum spend upfront. It's especially helpful for teams managing budgets across multiple chatbots or accounts.
Good to know
- Only resolved conversations count toward your spend cap.
- Only account owners can configure the spend cap from the Billing page.
- Only account owners receive spend alert emails.
Agent Traces help you understand how your AI agent arrived at a response by showing the steps it took behind the scenes. For supported messages, you can now see things like knowledge base searches, query expansion, date/time lookups, escalations, and custom action runs, all in the chat view.
How to use it:
- Open any conversation in your agent chats
- Find an AI message that includes trace data
- Click the Agent Trace section above the message bubble to expand it
What it’s useful for:
- Debugging why the agent answered a certain way
- Seeing which knowledge base searches were performed
- Reviewing which tools or actions were triggered
- Understanding escalations and other decision steps more clearly
Understanding how Helply AI Agent queries your articles will help you fine-tune training materials and increase the Agent's performance. 🚀.
You can now access Helply conversation data through the new External API, designed for analytics and reporting use cases. It allows you to fetch chats and their messages for specific agents, enabling use outside of Helply.
What’s special:
Provides direct access to full chat histories (including messages), with filtering by state and time range, pagination support, and an OpenAPI spec for easier integration.
Who it’s for:
Teams that want to build custom reports, dashboards, or internal data pipelines using Helply data.
Example:
Exporting conversation data into a BI tool to track support performance, trends, or agent activity over time.
Access & security:
Generate an External API key in
Settings → API
, and use it as a bearer token in the Authorization header. The key is only shown once — if lost, it must be regenerated, which will invalidate the previous key.
What’s included:
- Chat list and single chat retrieval
- Messages included in chat detail responses
- Filtering by state and time range
- Cursor pagination
- Rate limiting for secure access
- OpenAPI spec for easier client integration
For more information, see our Knowledge Base help page.