Our latest updates

Copy agents and databases to start faster

We added copying for agents and databases, a quality of life improvement that makes it much easier to build from work you have already done.

Start your next agent from one that already works

Many teams build agents in patterns. One agent watches a finance process, another watches the same process for a different team. One agent prepares a monthly report, another needs almost the same logic with a different data source or schedule.

You can now start from an existing agent instead of recreating the setup by hand.

That makes it faster to create variations, roll out a proven workflow to another part of the business, or experiment with a new version while keeping the original agent intact. The boring setup work is already done, so you can focus on the few things that need to be different.

Reuse database structure and data

Databases often become the working memory behind an agent. They hold cleaned records, snapshots, intermediate analysis, operating rules, or the structure for a recurring report.

You can now copy a database before making changes, or use an existing database as the base for a new workflow.

That is useful when you want a backup before trying a bigger change, when you want to test a different structure, or when a new customer or team needs the same kind of database with a few adjustments. You keep the structure and data you already trust, then work from there.

Major upgrades to workspace chat

We have rolled out a series of powerful upgrades to Workspace Chat to make it an even more capable and intuitive command center for your entire workspace.

File Attachments in Workspace Chat

You can now upload files and documents directly into your workspace chat sessions. Whether you want to share a CSV of customer records, a PDF guide, or a text file, the assistant can now load and inspect these attachments to answer your questions with full context. This brings workspace chat to full feature parity with agent and database chats.

Smarter, Faster Resource Discovery

We have completely overhauled how the assistant searches and interacts with your workspace resources. The assistant can now find and query groups, agents, logs, documents, connectors, databases, and users much more reliably. It retrieves exactly the data needed to answer your query, resulting in faster response times and more accurate answers.

Human-Readable Links

No more raw technical IDs cluttering your chat history. The assistant now automatically formats links to agents, groups, documents, databases, and logs using their human-readable names, summaries, or descriptions. Navigating from your chat conversation to the relevant resource is now completely seamless.

Safer Agent Execution

To prevent accidental runs, the assistant will now always ask for your explicit confirmation of the agent name and runtime arguments before executing any agent from the chat interface. You are always in control of when and how your automations run.

Shared reports are now easier to understand at a glance

Reports should not make people work to understand what changed.

UpdateMate agents can now create shared documents with richer visual reports, so the important parts of an update are easier to scan, explain, and act on.

Show the trend, not just the number

Line charts make it easier to see direction over time. Instead of reading a list of daily values, a teammate can immediately understand whether revenue is improving, ticket volume is rising, churn risk is stabilizing, or a campaign is losing momentum.

This matters most for recurring work. Weekly business reviews, investor updates, customer health summaries, and operational reports all become more useful when the reader can see the shape of the change, not just the final value.

Compare what matters side by side

Bar charts help agents turn comparisons into something people can read quickly. They are useful for showing channel performance, customer segments, product usage, pipeline stages, support categories, or cost breakdowns.

When a report is sent to a team, the reader should not have to calculate which category is leading or falling behind. The visual comparison makes the answer obvious.

Explain uncertainty clearly

Forecasts are rarely a single clean number. They usually come with a likely range, a best case, and a worst case.

UpdateMate documents can now show uncertainty bands, making forecasts easier to trust and discuss. A sales forecast can show the expected path and the confidence range around it. A cash forecast can show the likely runway while still making risk visible. A customer metric can show where the signal is strong and where the data is still uncertain.

This gives readers a more honest view of the situation. They can see both the recommendation and the confidence behind it.

Make stacked values easier to reason about

Some reports are about totals, but the total only makes sense when you can see what it is made of.

Stacked charts help agents show how different parts contribute to a whole: revenue by product line, support tickets by priority, pipeline by source, spend by channel, or usage by customer tier.

That makes it easier to answer the next question people usually ask: what is driving the change?

Share a polished report without rebuilding it elsewhere

The biggest improvement is not just that charts look better. It is that agents can include the visual context directly in the document they share.

That means a report can go straight from an agent run to a teammate, stakeholder, or customer without rebuilding the same information in a spreadsheet, slide deck, or dashboard. The recipient gets the explanation and the visual evidence together, in one place.

For teams using UpdateMate for recurring analysis, this makes agents better at producing finished work, not just raw answers.

Improved browser usage for agents

Agents are now better at working with websites and browser-based tools.

They can read the current page as clean text, making it easier to understand dashboards, reports, documents, and pages that do not have a simple API.

Browser snapshots are also more reliable. Instead of relying as much on fixed waits, agents now observe the page until it becomes stable before deciding what to click, read, or report back.

This makes browser-based agents easier to build and more dependable for everyday operational workflows.

Agent-ready workspace databases

UpdateMate now has workspace databases built for agents.

You can create structured SQL databases, inspect tables, ask questions through database chat, and let agents use the data as part of their workflows.

Databases include table previews, expandable rows, custom query views, descriptions, and snapshots before larger changes.

That means agents can work from live structured data instead of fragile spreadsheets or one-off exports.

Share richer documents and reports

UpdateMate documents are easier to share and read.

Agents can create documents and reports that include richer formatting, tables, charts, and HTML output.

Shared documents can be opened through a dedicated link, making it simpler to send agent results to teammates, stakeholders, or customers without copying everything into Slack or email.

This makes UpdateMate a better place for recurring reports, summaries, and longer analysis outputs.

Chats are now a workspace-wide command center

Chats are now a first-class workspace area in UpdateMate.

You can use workspace chat to ask questions across your agents, logs, documents, connectors, databases, groups, and users.

Chats can also stay attached to specific agents or databases, so the conversation keeps its context while you refine a process, debug a run, or inspect data.

Unread states and status updates make it easier to see when something needs your attention.

Add attachments to chat

You can now add attachments to chat messages.

This makes it easier to give agents the context they need without pasting long text or manually moving data around.

Attach files when you want an agent to review source material, use a document as context, or build a workflow around the data you already have.

Attachments make chat a better starting point for real work, not just short instructions.

Agents can work with browser-based tools

Agents can now work with more of the tools your business actually uses.

We added browser tooling and password connector support so agents can interact with systems that are harder to reach through a simple API key or OAuth flow.

This opens the door for workflows where agents need to read pages, navigate browser-based tools, or use authenticated systems as part of a process.

It makes UpdateMate useful for more internal operations work, especially when the tools involved were not built for automation.

A better connector experience

We have rebuilt the connector experience.

Connectors now have a dedicated page, clearer setup flows, better error messages, and support for more authentication types.

You can set up API key connectors, OAuth connectors, Intercom, Google services, and password-based connectors, then reuse them across agents.

Credentials stay stored securely and out of agent code and logs, while agents get the access they need to finish real work.

Clearer logs for every agent run

Agent logs are now much easier to read.

Logs show each run as a timeline with steps, live progress, generated documents, errors, and summaries of what happened.

Errors can include expandable details and a path back to the agent chat, so you can ask the agent to explain or fix the problem.

Logs also support markdown and richer output, which makes long-running or complex agents easier to monitor and debug.

Sign in with Google

You can now use Google Sign-In with UpdateMate.

This makes login and signup faster for teams that already use Google Workspace.

We also improved the authentication flow with clearer password and magic-link options, so it is easier to choose how you want to access your workspace.

Less account friction means you can get to building agents faster.

A new inbox for agent activity

UpdateMate now has an inbox for agent activity.

The inbox helps you see which conversations need attention, when an agent is thinking, and when something has failed.

Unread counters and real-time status updates make it easier to keep up with the work your agents are doing.

When an agent run fails, the inbox can bring that back to the right user so it does not disappear silently in the background.

More reliable long-running agents

We have made agent runs more reliable.

Long-running agent work now survives more backend restarts, stuck statuses are handled better, and agents can be stopped when they are taking too long or heading in the wrong direction.

When something fails, UpdateMate records the error in the log and gives you a path back to the agent chat so you can fix and rerun it.

This makes scheduled and background agents safer to use for real operational work.

Better agent building through chat

Agent building through chat is now much stronger.

You can ask what an agent does, request changes, add error handling, refine outputs, and keep iterating until the workflow behaves the way you want.

Under the hood, UpdateMate is better at planning changes, editing agent code safely, checking its work, and updating the agent description so it stays self-documenting.

This makes agents easier to maintain as your process changes.

Authenticated feed and PWA support

We added an authenticated feed view and PWA support.

This makes it easier to access agent outputs and document updates from inside the app, including on mobile.

The feed can be opened in the app layout, and mobile navigation has been improved so documents and updates are easier to review.

It is a small step toward making UpdateMate feel more like a daily operating layer for your business.

Organize agents into groups

You can now organize agents into groups.

Groups make it easier to keep agents separated by team, process, or area of work, like Sales, Marketing, Support, Finance, or Operations.

The sidebar supports collapsible groups and drag-and-drop organization, so your workspace stays manageable as you add more agents.

This is especially useful once agents become part of day-to-day operations across multiple teams.

Documents can include charts and structured reports

Agents can now create richer documents.

Documents can include structured reports, tables, and charts, making them more useful for dashboards, weekly updates, customer analysis, and internal summaries.

You can open documents from agent runs and share the output when teammates need to review what an agent produced.

This turns agents from simple chat responders into tools that can create reusable business artifacts.

Run agents on a schedule

Agents can now run automatically on a schedule.

You can run agents manually, or schedule them to run every minute, every hour, every day, every weekday, on specific days, or on monthly schedules.

This is what makes agents useful for recurring work like daily checks, weekly reports, data cleanup, monitoring, and operational handoffs.

Instead of remembering to run a process, you can let the agent run it for you.

Live status for long running reports

Sometimes loading and analyzing data can take some time.

But we humans are rarely patient when it comes to waiting.

We've added a live status for long-running reports so they are easier to follow.

This way, you can always see how far we are in the analysis.

ShopifyQL Examples for Automated Shopify Reporting

ShopifyQL examples are useful when you want automated Shopify reports without rebuilding the same analytics view by hand every week. UpdateMate can use ShopifyQL metrics inside AI agents so ecommerce teams can track revenue, refunds, products, conversion, and customer trends in recurring reports.

ShopifyQL is Shopify's query language for analytics and reporting. It looks familiar if you know SQL, but it has its own commerce-focused syntax, tables, operators, and reporting rules.

What is ShopifyQL?

ShopifyQL is a query language built by Shopify for analyzing store data. A ShopifyQL query usually starts with a data source, chooses the metrics or dimensions to show, then adds filters, time ranges, grouping, ordering, limits, and visualization options.

A simple pattern looks like this:

FROM sales
SHOW total_sales
SINCE startOfDay(-30d) UNTIL today

That kind of query can become a metric inside UpdateMate, so an AI agent can read it, compare it to previous periods, and explain what changed.

ShopifyQL tables

ShopifyQL queries use FROM to choose the source table or dataset. Common reporting areas include sales, orders, customers, sessions, products, and other Shopify analytics data that Shopify exposes through ShopifyQL.

Examples of report questions that map well to ShopifyQL tables:

• Sales: What was total sales this week, month, or quarter?
• Orders: How many orders were placed and how did average order value change?
• Products: Which products or variants sold the most?
• Customers: Which customer groups bought repeatedly?
• Sessions: Did traffic or conversion rate change?
• Refunds: Which products or channels are causing refund pressure?

When building reports, start with the business question first, then choose the ShopifyQL table, metrics, dimensions, and date range that answer it.

Does ShopifyQL support LIKE or ILIKE?

ShopifyQL is SQL-like, but it is not the same as SQL. In Shopify's current syntax reference, string matching is documented with operators such as STARTS WITH, ENDS WITH, and CONTAINS, not standard SQL LIKE or PostgreSQL-style ILIKE.

So if you are searching for a ShopifyQL LIKE operator, the practical answer is: use ShopifyQL's documented string matching operators instead of assuming LIKE works.

Examples of the intent:

WHERE product_title CONTAINS 'shirt'

WHERE customer_email ENDS WITH '@example.com'

If you need case-insensitive behavior like ILIKE, test the exact query in Shopify's ShopifyQL editor or API for your field and data. Do not assume SQL ILIKE syntax is available just because ShopifyQL looks similar to SQL.

ShopifyQL examples for ecommerce reporting

Here are practical ShopifyQL reporting ideas that work well as automated metrics for an ecommerce team.

Total sales over the last 30 days:

FROM sales
SHOW total_sales
SINCE startOfDay(-30d) UNTIL today

Sales by product title:

FROM sales
SHOW total_sales
GROUP BY product_title
SINCE startOfDay(-30d) UNTIL today
ORDER BY total_sales DESC
LIMIT 10

Sales over time:

FROM sales
SHOW total_sales
TIMESERIES day
SINCE startOfDay(-30d) UNTIL today

These examples are meant as starting points. The exact fields available can depend on the ShopifyQL dataset and analytics access for the store.

Revenue, refunds, and product performance examples

UpdateMate works best when ShopifyQL metrics map to recurring business questions, not isolated charts.

Useful ecommerce report questions include:

• Which products drove the most revenue last week?
• Which products declined fastest compared with the previous period?
• Are refunds increasing for a specific product, vendor, or campaign?
• Did average order value move up or down?
• Which products are growing but still low in absolute revenue?
• Which customer or order segments need attention?

An AI agent can combine ShopifyQL outputs with context from ad platforms, inventory sheets, support tickets, or campaign notes to explain why the numbers changed.

How UpdateMate turns ShopifyQL into recurring reports

UpdateMate can use ShopifyQL to power automated Shopify reporting workflows.

A recurring workflow can:

1. Run ShopifyQL metrics on a schedule.
2. Compare sales, product, refund, and customer metrics with the previous period.
3. Detect meaningful changes, anomalies, and trend breaks.
4. Pull supporting context from other connected tools.
5. Write a weekly ecommerce report in plain language.
6. Send the report to Slack, email, Notion, Google Docs, or another destination.

This turns ShopifyQL from a query you run manually into a repeatable reporting system for ecommerce operators, founders, marketers, and finance teams.

GPT-4o mini in the UpdateMate Free Plan

GPT-4o mini is available in the UpdateMate free plan, so you can test practical AI agent workflows before upgrading. This page is about GPT-4o mini free plan availability inside UpdateMate, not a general OpenAI model comparison.

Previously, only gpt-3.5-turbo was available for free users. Adding GPT-4o mini in UpdateMate makes the free AI agent plan more useful for testing real business workflows with larger prompts, files, and structured context.

What GPT-4o mini can do in UpdateMate

In UpdateMate, GPT-4o mini is useful for lightweight agents that need to read business context, summarize information, classify records, or draft recurring updates.

For example, you can use it to:

• Summarize support tickets.
• Draft weekly reports.
• Classify CRM records.
• Turn spreadsheet rows into action items.
• Summarize notes from sales or customer success calls.
• Draft status updates from existing data.
• Test agent instructions before moving a workflow into production.

What is included in the free plan

The UpdateMate free plan lets you try GPT-4o mini with real AI agent workflows, so you can see whether an automation idea is useful before committing to a paid setup.

That means you can:

• Create an UpdateMate agent.
• Write instructions for a workflow.
• Test the workflow with your own business context.
• Review the output before deciding whether to automate it more deeply.

A context window is how much data a model can work on at once. The larger the context window, the more data a model can analyze in one request.

Tokens are how generative AI breaks down data. For text, a token is roughly a short word or word fragment. Exact token behavior varies by model.

Example free-plan AI agent workflows

Here are practical workflows to test with GPT-4o mini in UpdateMate:

• Summarize the last 20 Zendesk or Intercom conversations and list the top customer issues.
• Draft a weekly marketing report from campaign metrics in a spreadsheet.
• Classify HubSpot contacts by segment, lead quality, or missing CRM fields.
• Turn Google Sheets rows into a prioritized task list.
• Summarize customer feedback into product themes.
• Draft a founder or team update from bullet points and metrics.

These are good free-plan workflows because they are useful, repeatable, and easy to review.

When to use a larger model

GPT-4o mini is a good starting point for many UpdateMate agents, but some workflows may need a larger model or a paid plan.

Consider upgrading when:

• The agent needs deeper reasoning over messy or high-stakes context.
• The workflow runs frequently or needs more production capacity.
• The output needs more nuance, stronger writing, or more reliable judgment.
• The agent needs to process larger or more complex documents.

Try GPT-4o mini in UpdateMate's free plan.

HubSpot support conversations

Now you can get your AI Agents connected to all your support conversations in HubSpot with a single click.

This means that you can now get a full 360° view of what your customers are saying in both sales and support.

We have connected some of the most powerful support insights here.

Quick start with our most popular agents

Building an AI Agent from scratch can be a tall order—especially when we're at the cutting edge of human capabilities. Many of our customers are first movers, if not the very first, to use AI Agents to solve real problems.

But sometimes, it's nice to just get going instead of being first at everything 😅

So, we've created some great templates to start from when creating a new agent, based on what has been most popular among our customers.

One-click AI Agent Templates:

• Won deals analysis: Why do customers choose us? Analyze your won deals in your CRM to learn what makes your customers choose you.
• Lost deals analysis: Why do customers not choose us? Analyze your lost deals in your CRM to learn from your mistakes.
• Why do customers need support?: Why do customers reach out for support, and what should you fix to reduce your support load?
• Why do customers request refunds? Refunds directly impact your bottom line. Analyze why your customers request them to minimize your losses.
• C-level direct customer feedback: Always be aligned with your customers. Get weekly updates on what customers love about you and what you should change immediately.
• Team praise: Go home on a high note with a praise bot that shares your customers' best praise with your team at the end of the day.

You can also (as always) just build your own from scratch.

These are just the first ones to start out with. We will add more as we learn with you.

Filter Hubspot deals by real activity

Some of your most valuable data is the conversations your sales reps have with customers.

So we have added features to fine tune the data your agents work for.

Filter deals based on activity

• Get deals with calls made within the last 1, 2, 3, 5, 7, 14 and 30 days.
• Get deals with emails made within the last 1, 2, 3, 5, 7, 14 and 30 days.
• Get deals with meetings made within the last 1, 2, 3, 5, 7, 14 and 30 days.
• Get deals with notes made within the last 1, 2, 3, 5, 7, 14 and 30 days.
• Get deals with tasks made within the last 1, 2, 3, 5, 7, 14 and 30 days.

You can also choose to include all or none of the above and filter out deals with no activity meeting that filter.

This allows you to get good data for analytics of calls, emails and other conversations on a daily, weekly or monthly basis.

Filter deals based on pipelines

You can also filter deals on deal stages across all of your pipelines.

This allows you to eg take all won deals in different pipelines and analyse why customers bought from you.

Claude 3.5 Sonnet now avaliable

We have added Anthropics Claude 3.5 Sonnet model to UpdateMate.ai

Claude 3.5 Sonnet is a great model for working on code and structured data.

It comes with a 200k context window so it can be a great replacement for OpenAIs GPT-4o or GPT-4-Turbo models (that only have 128k context window) without sacrificing quality.

Share your updates with the world

We just made any update from UpdateMate.ai shareble with everyone.

Share with anyone ad-hoc

Any update can now be shared with a shareable link by just clicking the share button at the top right.

Easier to share big report in Slack or e-mail

Sometimes a long report can be hard to read in slack or email. It just gets overwhelming.

Now you can choose if a Agent should send everything directly or if the update should be shared with a link and then only a small summary in the email or Slack message.

How to Trigger AI Agents With Webhooks

You can trigger AI agents with webhooks when a workflow needs to run in real time instead of on a weekly or daily schedule. In UpdateMate, an AI agent can receive an inbound webhook, analyze the payload, take the next step, and send its output to another webhook, Slack channel, email, database, or API endpoint.

This makes AI agents useful for support alerts, phone agent notifications, lead routing, incident summaries, customer updates, and any workflow where another system needs to wake an agent up immediately.

What is a webhook?

A webhook is a message that one application sends to another application when something happens. Instead of asking a system for updates every few minutes, the system sends a real-time notification to a webhook URL.

Examples:

• A support platform sends a webhook when a customer opens a high-priority ticket.
• An AI phone agent sends a webhook after a call ends.
• A form tool sends a webhook when a new demo request arrives.
• A billing system sends a webhook when a payment fails.
• A monitoring tool sends a webhook when an incident starts.

The webhook payload usually contains structured data, often JSON, that describes what happened.

How webhook-triggered AI agents work

Webhook-triggered AI agents use the webhook payload as the starting context for a run.

A typical flow looks like this:

1. An external system sends a webhook to a unique UpdateMate agent URL.
2. UpdateMate starts the AI agent immediately.
3. The agent reads the payload, connected tools, and any relevant historical context.
4. The agent analyzes the situation and decides what output is needed.
5. The agent sends a response, alert, summary, database update, or outbound webhook notification.

This lets agents act as real-time workflow middleware. They can receive events from one system, understand what happened, and send a useful output to another system.

Inbound webhook triggers

An inbound webhook trigger lets another system start an UpdateMate AI agent.

Use inbound webhooks when you want an agent to run after events like:

• A support ticket is created or escalated.
• A customer conversation becomes angry or urgent.
• An AI phone agent finishes a call and sends the transcript.
• A new lead submits a form.
• A deal changes stage in a CRM.
• A product usage event crosses a threshold.
• A monitoring tool detects an outage.

In UpdateMate, the agent gets a unique webhook URL. You place that URL in the system sending the event, then choose what the agent should do with the incoming payload.

Outbound webhook notifications

An outbound webhook lets an AI agent send data to another system after it finishes its work.

Use outbound webhooks when another workflow needs the agent's result, such as:

• Sending a structured summary back to a CRM.
• Posting a risk score to a customer success platform.
• Creating a task in a project management tool.
• Triggering a follow-up workflow in Zapier, Make, n8n, or a custom backend.
• Sending JSON to an internal API endpoint.

You can ask the AI agent to format its output as plain text, JSON, XML, Markdown, or another structure that the receiving system expects.

Example: send a Slack alert from a webhook

Imagine a support system sends a webhook whenever a VIP customer opens a new ticket.

UpdateMate can receive the webhook, read the ticket details, check customer context, and send a Slack alert like:

VIP support alert: Acme opened a billing ticket with negative sentiment. ARR: $84,000. Open renewal opportunity: yes. Suggested next step: have the account owner reply within 15 minutes and check the latest invoice notes.

This is more useful than forwarding the raw webhook because the agent turns the event into context, priority, and recommended action.

Example: return JSON from an AI agent

Webhook-triggered AI agents can also return structured data to another system.

Example JSON output:

{
  "priority": "high",
  "sentiment": "negative",
  "customer_tier": "enterprise",
  "recommended_owner": "account_manager",
  "summary": "Customer is frustrated about a billing issue before renewal.",
  "next_action": "Escalate to account owner and billing specialist."
}

This is useful when your backend, CRM, or automation tool needs machine-readable output rather than a human-written message.

Security checklist for AI agent webhooks

Webhooks can trigger real workflows, so treat them as production entry points.

Before using AI agent webhooks broadly, check:

• Use unique webhook URLs for each agent or workflow.
• Keep webhook URLs private and do not paste them into public docs.
• Validate the sending system when possible with signatures, tokens, or allowlists.
• Avoid sending unnecessary personal or sensitive data in webhook payloads.
• Log webhook runs so you can inspect what triggered an agent and what it returned.
• Use human review for workflows that send external messages, change customer records, or affect money.
• Define fallback behavior for malformed payloads, missing fields, and duplicate events.

With UpdateMate, webhooks let AI agents move from scheduled reporting into real-time operations. The agent can be triggered the moment something happens, understand the event, and send the right output to the right place.

HubSpot integration

We just added HubSpot to our list of official integration.

This enables you to export your deals from HubSpot with a few clicks.

Features

• Export deals by pipeline and stage.
• Export custom deal properties.
• Export deal calls.
• Export deal emails.
• Export deal notes.
• Export deal meeting notes.
• Export deal tasks.

GPT-4o-mini now avaliable

We have added OpenAIs GPT-40-mini model to UpdateMate.ai

GPT-40-mini is OpenAIs successor to GPT-3.5-turbo and the scaled down model of their flagship model GPT-4o-mini.

It comes with a 128k context window so it can be a great replacement for GPT-3.5-turbo that only had a 16k context window.

Also it comes with much improved language and reasoning features and at the lowest price we have seen for a LLM to date.

ProductBoard integration avaliable

You can now pull all your product feedback from ProductBoard directly into your AI agents.

Use this to keep your team up to date with product requests or map requests against support queries to identify what goes unreported.

Google Gemini models avaliable

You can now use Googles Gemini Flash and Pro models in your AI agents.

Why more models do you ask? Well where Gemini models really shine is their 1 million token context window!

That is 8x more than any of OpenAIs model.

These models are great for analysing huge amounts of customer conversations.

Hello, World

We are UpdateMate.AI - a platform for getting AI Agents to do your reporting for you.

Our goal is to remove the most boring part of work - reporting and at the same time give people better insights into their business and customers.

Lets go!