Macha March 2026 Update: Triggers, Sub-Agents, Freshdesk & Shopify

Quick reminder on what Macha is, because it frames every item below: Macha is an AI agent layer that sits on top of your existing helpdesk — Zendesk, Freshdesk, Gorgias, Front — and connects your commerce, knowledge, and comms tools so agents can resolve, triage, and route tickets while taking real actions through tools. March's theme was autonomy: less "answer when asked," more "do the work when the event happens."

The headline: agents that run themselves

Triggers — agents that fire on events (March 1)

The trigger system is the biggest shift of the month. An agent no longer needs a human in the loop to start working. You point a trigger at an external event and the agent runs autonomously when that event fires.

Out of the box, triggers support Zendesk ticket events, Freshdesk updates, Slack mentions, and custom webhooks — and the custom-webhook path means any system that can send an HTTP POST can wake an agent. When you create a custom webhook trigger, Macha generates a unique URL with a signing secret so you can verify the payload is genuinely from your source.

Two details make this production-safe rather than a foot-gun:

• Debounce. Real systems emit bursts of events — a flurry of ticket updates in a few seconds. A configurable debounce interval batches rapid events so the agent fires once on the settled state instead of stampeding.
• Scoped triggers per agent. Each trigger is attached to a specific agent, so the right specialist responds to the right event.

This is the foundation the rest of the month builds on. See the triggers documentation for the full event list and payload format, or the deeper walkthrough on automating agents with triggers.

Scheduled triggers — cron for your agents (March 22)

Event-driven is half the story; the other half is time-driven. Scheduled triggers let an agent run automatically on a recurring interval — think a 9am sweep of overnight tickets, an hourly stale-conversation check, or a nightly digest.

The guardrails here are deliberately conservative:

• Minimum 1-hour intervals, with daily caps and concurrency limits so a schedule can't quietly burn through your credit balance.
• Auto-disable on failure — a trigger that hits 5 consecutive errors disables itself, so a broken integration doesn't loop forever.
• Plan limits: Professional includes 3 scheduled triggers; Enterprise includes 20. (See the pricing page for current plan details.)

Because every agent run consumes credits, the daily cap and concurrency limit aren't red tape — they're the difference between a useful nightly job and a runaway loop. Set the cap to the worst case you'd tolerate, not the best case you hope for. For cron patterns and recurring-run recipes, see running agents on a schedule.

Sub-agents — delegation between agents (March 8)

The third piece of the autonomy story is sub-agents: an agent can now call another agent as a tool. That enables hierarchical, specialist delegation — a front-line triage agent can hand a billing question to a billing specialist, or a refund flow to a Shopify-connected agent, instead of trying to be expert at everything itself.

In practice this keeps each agent's instructions and tool set tight. A lean orchestrator that knows who to delegate to beats one bloated agent with thirty tools and a 2,000-word prompt. Pair sub-agents with triggers and you get a clean pattern: a trigger wakes the orchestrator, the orchestrator routes to the right specialist, the specialist acts. The sub-agents docs cover how delegation and tool-name disambiguation work, and our guide to multi-agent delegation with sub-agents shows when the pattern pays off.

New connectors: Freshdesk and Shopify

Freshdesk connector — 15 tools (March 15)

Macha started Zendesk-first, but plenty of teams run Freshdesk. The March 15 connector brings full ticket management: get and search tickets, add replies and private notes, update priority / status / tags / custom fields, assign tickets, search contacts, and browse knowledge-base articles — 15 tools in all.

Two parts make it genuinely useful rather than a thin API wrapper:

• Attachment extraction. Agents can read and extract text from PDF, DOCX, XLSX, CSV, and TXT attachments directly off Freshdesk tickets — so a customer's attached invoice or log file becomes context the agent can reason over.
• Webhook triggers. Combined with March 1's trigger system, agents can respond automatically on ticket creation and updates.

If you run Freshdesk, this is the connector that makes Macha a same-day fit. Browse the full integrations library for the rest.

Shopify connector — commerce actions, including refunds (March 22)

For ecommerce support, the helpdesk is only half the context — the store is the other half. The Shopify connector connects via OAuth and gives agents 7 tools: search_products, get_order, lookup_customer, get_customer_orders, search_orders, get_discounts, and create_refund.

That last one matters: create_refund is a write action gated behind confirmation, consistent with how Macha treats every state-changing tool. The agent proposes the refund; a human (or an explicit confirmation step) approves before money moves.

This is squarely aimed at the highest-volume ecommerce tickets. A 2026 Shopify Plus merchant study cited across the industry found roughly 67% of high-volume DTC support tickets fall into three buckets — order tracking, returns, and exchanges (Fini Labs). With Shopify order lookups plus the helpdesk connector, a Macha agent can resolve those end-to-end. There's also a pre-configured Shopify support agent on a demo account so you can see the flow before connecting your own store.

A note on positioning: Macha isn't trying to be Gorgias or another Shopify-native helpdesk — those own the inbox. Macha is the agent layer that plugs into whichever helpdesk you already use and adds the Shopify context and actions on top.

Tooling, onboarding, and privacy

AI Tool Builder — connect any REST API by describing it (March 30)

The AI Tool Builder is the quiet power feature of the month. Instead of hand-defining endpoints, you describe the API you want to connect, paste your credentials, and the AI discovers endpoints, tests them, and creates the tools for you. It works with any REST API authenticated by API key, bearer token, or basic auth.

The thoughtful bits:

• Auto-grouping — generated tools are bundled under a named connector with an emoji icon and brand color.
• Smart auth handling — credentials captured from a successful test request are applied to every tool in the session.
• Auto-detect write operations — POST/PUT/PATCH/DELETE methods are automatically flagged as write operations requiring confirmation, so destructive calls don't run silently.

Custom Tools also got polish the same day: masked credential hints in the edit modal (verify which key is set without exposing the secret), inline delete from the tools list, and automatic stripping of empty optional parameters from request URLs (which quietly fixes errors from APIs that reject empty query values). Professional includes 5 custom API tools; Enterprise includes 20.

Onboarding — set up in under two minutes (March 8)

New workspaces now get a guided 3-step onboarding: select your tools, pick from recommended agent templates, and connect your integrations. The goal is a working agent in under two minutes instead of a blank canvas.

Multi-instance connectors (March 8)

Agencies and multi-brand teams asked for this: you can now connect multiple instances of the same connector — two Zendesk accounts, three Stripe accounts — each with its own credentials and name. When duplicate connector types exist, Macha auto-disambiguates by suffixing tool names and tagging descriptions with the account name, so the model always knows which instance to call.

Chat attachments (March 1)

Alongside triggers, March 1 added drag-and-drop attachments in chat: images, documents (PDF, DOCX, CSV, XLSX), and audio. Text is automatically extracted from documents into the conversation context, and images are sent as vision content for models that support it.

Privacy & data management (March 30)

The month closed with a privacy pass: a new Privacy settings page to download your data, manage account deletion, and reach every policy in one place. Data export produces a complete JSON copy of your conversations, agents, connectors, and tools. There's a cookie-consent banner on all pages, and automatic cleanup — conversations auto-delete after 45 days of inactivity, and chat attachments expire automatically.

A worked pattern that uses March end-to-end

Here's how the month's pieces compose into one workflow, which wasn't possible in February:

1. A trigger fires on a new Freshdesk ticket (Mar 1 + Mar 15).
2. A lean orchestrator agent reads the ticket and any attached files (Freshdesk attachment extraction).
3. It delegates an order question to a Shopify sub-agent (Mar 8 + Mar 22).
4. The sub-agent looks up the order, and if a refund is warranted, proposes create_refund — gated behind confirmation (Mar 22).
5. A separate scheduled trigger sweeps any still-open tickets each morning and posts a digest (Mar 22).

Every step is event- or time-driven, scoped to a specialist, and metered against credits with caps. That's the shape of "agents that run themselves."

Watch-outs and when not to lean on this

March added a lot of autonomy, and autonomy needs guardrails. A few honest caveats:

• Scheduled triggers spend credits unattended. That's the whole point, but it also means a too-loose daily cap on a chatty agent can surprise you. Start with conservative caps and a low concurrency limit, then raise them once you've watched a few runs. Credits are charged per AI action (0.5–9 credits by model, default GPT-5.4 Mini = 1), not per resolution — outcomes vary, so meter the actions, not the hopes.
• Triggers without good instructions are just faster mistakes. An agent that responds autonomously to every ticket needs tight scoping and a clear "when in doubt, hand off to a human" rule. Don't point a trigger at a brand-new, untested agent.
• The AI Tool Builder is powerful, so confirmation gates matter. Auto-flagging write methods is a safety net, not a substitute for reviewing what the generated tools can do before you let an agent call them on live data.
• Sub-agents add latency and cost. Each delegated call is another model run. Use them where specialization genuinely pays; a single well-scoped agent is cheaper for simple queues.
• Shopify create_refund moves real money. Keep it behind confirmation. The convenience of one-click refunds is not worth an un-gated agent issuing them.

If your queue is small, simple, and well-served by reply-only agents, you may not need triggers or schedules yet — and that's fine. These features earn their keep at volume and with workflows that have clear hand-off rules.

FAQ

What are the headline features in Macha's March 2026 update? Autonomous triggers (agents fire on Zendesk/Freshdesk/Slack/custom-webhook events), scheduled triggers (cron-style recurring runs), sub-agents (agents delegating to other agents), a Freshdesk connector (15 tools), a Shopify connector (7 tools including gated refunds), an AI Tool Builder for any REST API, guided onboarding, multi-instance connectors, chat attachments, and a privacy/data-management page.

How often can scheduled triggers run? Minimum 1-hour intervals, with configurable daily caps and concurrency limits. A trigger auto-disables after 5 consecutive errors. Professional includes 3 scheduled triggers; Enterprise includes 20.

Does Macha now support Freshdesk as well as Zendesk? Yes. The March 15 Freshdesk connector adds 15 ticket-management tools plus attachment text extraction and webhook triggers — Macha remains an agent layer on top of your helpdesk, now supporting Freshdesk natively alongside Zendesk, Gorgias, and Front.

Can a Macha agent issue Shopify refunds? Yes, via the create_refund tool — but it's a write action gated behind confirmation, so a human approves before any refund is processed.

Do I need to write code to connect a custom API? No. The AI Tool Builder lets you describe the API and paste credentials; it discovers and tests endpoints and creates the tools, auto-flagging write operations for confirmation. Professional includes 5 custom API tools; Enterprise includes 20.

Try it

If you've been waiting for agents that act on their own — fire on events, run on a schedule, delegate, and reach into Freshdesk and Shopify — March is the release to start a 7-day free trial, no credit card required on. Connect your helpdesk, point a trigger at it, and watch an agent do the work. Read the triggers docs and sub-agents docs for setup, or browse the Macha blog for deeper walkthroughs of each feature.

Written by Abbas (Customer Support & AI, Macha) · Reviewed by Ankeet Guha (Co-founder & CTO) · Published 2026-06-24 · Last updated 2026-06-24.