Settings

Pick the language you work in (primary), the languages you read natively (secondary), and the language you want the Atender UI to appear in.

A single toggle in Settings → Translation turns auto-translation on for the whole tenant. Each agent then sets their own language preferences.

The five things to check when an agent says messages aren't being translated, in order from most likely to least.

Auto-translation lets agents reply in their own language while customers write in theirs. The customer never sees the translation seam.

Every action a tool dock can trigger back in Atender — add a tag, assign, resolve, snooze, send a message — and the conversation context the dock receives in return.

Embed your internal order admin page as a tool dock so agents can lookup, refund, and resolve orders without leaving the conversation.

Pick a pre-built integration from the marketplace and walk through its guided OAuth setup — most one-click integrations land in your Toolbox in under a minute.

Embed any web URL as a tool dock — your CRM, an internal admin page, a custom dashboard — with optional OAuth and a single-letter shortcut.

The Toolbox brings external tools — ClickUp, your CRM, internal admin pages, anything — into the conversation view as slide-out panels, so agents never tab-switch.

Walk through the Staged Changes tab — read a proposal's diff, decide to approve, reject, or dismiss it, and see the change reflected in your stack.

Self-Learning only learns from feedback your team provides on your tenant's conversations. Your data never trains models for other tenants.

Self-Learning collects flagged AI responses, groups them into themes, proposes concrete fixes, and lets you approve and deploy improvements with a click.

Security levels 0–3 control how much oversight a capability gets — from silent execution all the way up to supervisor approval. Plus identity verification, redaction, and consent tracking.

Every node you can place on the visual canvas: what it does, when to use it, and what it produces for downstream nodes.

Every authentication type and auth mode a capability supports — when to use each, and what the difference between auth type and auth mode actually is.

Build a capability that takes an order number, calls your store's API, and returns the status, expected delivery, and tracking link to the AI specialist.

Run a capability against a real endpoint with sample inputs, inspect the response, and flip it from Draft to Published when everything looks right.

Build a new capability from scratch — pick a build mode, configure the API, define inputs and outputs, save as a Draft.

Capabilities are the actions your AI agents can take — looking up an order, processing a refund, calling any external API. Without them, agents can only talk.

The six personality presets a stack can adopt — and the underlying tone dimensions (formality, directness, warmth, confidence, humor, emoji) each one sets.

The state machine the orchestrator runs over every conversation in a stack — the states, the transitions, and what triggers each one.

The four background infrastructure agents that run on every Agent Stack — conversation summary, sentiment & risk, handoff pack, missing info — and the settings each one exposes.

The three handover modes a stack can be configured with — never, by_description, explicit_team — and what each one means for routing handed-off conversations to humans.

Configure handover so the AI gathers everything a human needs first — account email, order ID, sentiment context — and hands off cleanly to a single team during business hours, with email follow-up after hours.

Build a basic two-specialist Agent Stack — one for billing questions, one for technical issues — and hand off to the right team when a human is needed. Real, copy-paste configuration.

Use the built-in Testing tab to chat with your Agent Stack as a customer, isolate individual specialists, simulate identity verification, and watch the routing decisions in real time.

Pick a persona preset and tune the tone dimensions for your Agent Stack — the voice, formality, warmth, and humor that every specialist in the stack will share.

Configure which Knowledge Base categories and Handbook procedures a specialist can read. Tighter scope means faster, more accurate retrieval and fewer wrong answers.

Flag any AI reply that missed the mark, give feedback in plain language, and let Atender propose concrete edits — system prompt changes, routing tweaks — that you can apply with one click or revert.

Take an Agent Stack out of service temporarily by disabling it, or remove it permanently from the Danger Zone tab. Disabled stacks stop receiving new conversations but keep their config.

Most specialists route correctly from their Scope text alone. When the router keeps mis-routing, append explicit routing rules from the Orchestrator config or sharpen the Scope itself.

Create your first Agent Stack — the AI customer service team that handles inbound conversations. The guided wizard sets up the stack, suggests specialists, and wires up knowledge in five short steps.

Turn handover on, choose a destination team, set the prerequisites the AI must collect first, and pick what happens outside opening hours — all from the Handover tab.

Add a new specialist agent to an existing Agent Stack — give it a name, a scope, and instructions, then wire knowledge and capabilities.

Attach a capability to a specialist so it can act in your other systems — look up an order, process a refund, check delivery — and pick the security level that decides when the action is available.

When the router picks the wrong specialist, three things to check first — Scope text, orchestrator routing rules, and whether the orchestrator re-routed mid-conversation.

When an Agent Stack stops replying or never replies in the first place, work through these checks in order — channel wiring, stack enabled, specialist enabled, capabilities failing silently.

Specialists are the experts inside an Agent Stack. Each one is a focused AI agent with its own scope, instructions, knowledge access, and capabilities — and the more specific each one is, the better the whole stack performs.

An Agent Stack is your AI customer service team — a router, an orchestrator, and a set of specialist agents that handle conversations autonomously and hand off to humans when needed.

An Agent Stack is built from three layers — a router, an orchestrator, and one or more specialist agents — plus a set of background infrastructure agents that handle summaries, sentiment, and handoff prep.

The orchestrator is the agent that watches every conversation in a stack — it detects stalls, re-routes when the wrong specialist was picked, and triggers handover when the situation needs a human.

Handover is the boundary between the Agent Stack and your team. Configure when handover happens, what information must be collected first, and which team picks up — all from the Handover tab.

Every module that can appear in the Sidekick panel — summary, suggested answer, insights, knowledge articles, handbook articles, relevant links — what each does, when it's useful, and the data sources it draws from.

Three methods Sidekick can use to detect structured data in messages — pattern builder, AI detection, and raw regex. When each fits, plus the pre-built templates that ship with Atender.

When Sidekick detects that a customer is describing an active incident, a hint card surfaces above the panel with two actions — link & subscribe (auto-notifies the customer when the incident resolves) and insert reply (templated acknowledgment).

Build a Sidekick detection rule that extracts structured data from messages and writes it to a custom field. Walks through the pattern-builder method with a worked Order Number example.

Choose which Sidekick modules show up in the agent panel and the order they appear in. Plus the global toggles in Settings → Sidekick → General — incident discovery, inline highlights, auto-save, confidence threshold.

Sidekick is the conversation-intelligence layer that runs over every conversation. For agents, it's a slide-out panel with summaries, suggested responses, and matching knowledge. For the system, it runs background workers that auto-tag, extract structured data, and detect incident matches.

Every scope an API key can be granted, grouped by surface area. Plus the pre-built bundles (basic, integration, full_access) that cover common integration patterns.

Four rate-limit tiers — starter, professional, enterprise, custom — with per-minute, per-day, and burst caps. What 429 responses mean, how the sliding window works, and when to upgrade tiers.

Replace an API key without downtime — generate the replacement, update consumers, then revoke the old. Plus the emergency rotation flow when a key is leaked.

Watch per-key usage — request counts, error rates, last-used timestamps. Where the dashboards are, what to look at, and what alerting setup makes sense for production keys.

Create a new API key with the right scopes and rate-limit tier. The key is shown once at creation; copy it to a secrets manager immediately. Walks through a knowledge:read+write key for a doc-as-code pipeline.

API keys (sa_live_* / sa_test_*) authenticate external systems to Atender's tenant API. Each key has scopes that gate which endpoints it can hit, a rate-limit tier, and an environment (production vs sandbox). Used for integrations, scripts, and the API channel.

Every event Atender can trigger an automation on, grouped by the type of context it gives the rule (conversation, tag, case, SLA, schedule, business hours).

Every condition field and operator available when narrowing an automation rule, with notes on AND/OR groups and how conditions relate to branches.

How an automation runs different action sequences depending on what's true — Always, If, Else if, and Else branches.

Every action an automation can run, grouped by category — communication, conversation lifecycle, cases, external (webhook), and flow control (wait).

When a case closes with a specific tag (e.g. VIP-Refund), POST a webhook to a downstream system with the case ID, contact, and any relevant custom fields. The universal escape hatch for systems Atender doesn't natively integrate with.

When a new conversation comes in from a VIP contact, assign it to the Premium customer service team and tag it for visibility. Uses customer level on the contact to discriminate.

When a conversation's SLA is at risk, reassign it to a senior agent, ping the leads channel in Slack, and tag it as Escalated. One rule, three coordinated actions.

Three days after a conversation is marked Done, send the caller an Anything Else? template. Catches lingering follow-ups before they become a new ticket.

Three hours after a conversation is resolved, send a CSAT survey to the caller. Throttled to once per contact per week to avoid survey fatigue.

Auto-reply with a confirmation when a new email lands, in the inbox's own language, using an Email Studio template. Atender doesn't send confirmations by default — this rule is how you opt in.

Apply a tag automatically when an inbound message contains specific keywords. The classic example: tag anything mentioning 'refund' or 'return' as a Returns case.

Every day at 02:00, find Done conversations that have been untouched for 14 days and archive them. Keeps the inbox tidy without manual sweeping.

Describe what you want in plain English and let Atender draft an automation rule for you. Covers good prompt patterns, review steps, and the Explain Rule tool.

Rate-limit how often an automation fires per conversation, per contact, or globally. Stops chatty triggers from spamming customers or downstream systems.

Use Simulate Execution to see which conversations a rule would match — without sending any messages or running any actions — before you turn it on.

Restrict an automation to fire only inside or only outside business hours, using your workspace's opening hours configuration.

Use Manual Run to apply a rule to existing conversations — useful for one-off backfills, clean-up sweeps, and bulk re-tagging. Always simulates before executing.

Find why a rule isn't firing or why an action failed. Manual Executions logs every run with trigger, branch, action results, and full input/output payloads.

Build an automation in the manual builder — pick a trigger, add conditions, set up branches, drop in actions, and turn it on.

Automations watch your inbox for events, check whether the situation matches criteria you set, then run actions — assign, tag, reply, escalate — without an agent doing it by hand.

Turn on conversation IDs, pick a numbering strategy, add a prefix and date component, choose where the ID is displayed, and optionally override the prefix per channel or team.

An optional, configurable customer-facing identifier for conversations — sequential or random, with prefixes, dates, and per-channel or per-team overrides.

Configure how long a conversation sits in Done before Atender archives it. Timing is measured in business hours per the team's opening hours, not wall-clock.

Use the Mark as done button (or the D shortcut) to close out a conversation once the work is finished. Done conversations can still be reopened by a customer reply.

A customer reply on a Done conversation reopens it to Active. A customer reply on an Archived conversation does not — it creates a new conversation instead.

Conversation Resolution is how Atender treats a conversation as finished — the agent marks it Done, and Atender later archives it. Done is reopenable; Archived is terminal.

A repeatable pattern: host your help center at help.example.com and your status page at status.example.com under the same root domain, both managed in Custom Domains.

A custom domain stuck on Awaiting DNS, Issuing certificate, or Failed almost always traces back to a DNS issue. Here's how to find which one.

Map a hostname like help.example.com or status.example.com to your Atender Knowledge Base or Status page, with automatic TLS.

Custom Domains covers the Knowledge Base portal and the Status page. The agent app, Web Chat widget, and email sending domain are configured elsewhere or not at all.

Custom Domains let you serve your Atender Knowledge Base and Status page from your own hostnames — help.example.com instead of the default tenant URL. TLS is automatic.

Every custom field type, what it stores, when to use it, and the gotchas. Picking right the first time saves a lot of cleanup later.

A complete worked example for e-commerce support: order number, order status, fulfillment center, and refund amount custom fields, plus Sidekick auto-population and one Automation that ties them to routing.

Reference a custom field's value as a condition (route by tier, escalate by severity) or write to it as an action. The integration that makes custom fields more than just display data.

Split a busy contact panel into separate tabs (Account, Billing, History, Internal). The next layer of organization above sections — for tenants with 30+ contact custom fields.

Group related custom fields under named sections so the contact / conversation / case panel stays scannable as the field count grows. Required reading before adding more than 5–6 fields to any entity.

Add a new custom field to a contact, conversation, or case. Pick a name, type, optional default, and set whether it's required.

Configure Sidekick to scan inbound messages for patterns — order numbers, serial numbers, license keys — and write the extracted value into a custom field. Saves agents the copy-paste.

Yes, but the data goes with it. Deletion is hard-delete by design — Atender shows the affected record count first, and your real options are usually to migrate the data or stop populating the field.

Custom Fields extend Atender's data model with the information that matters to your business — order numbers on conversations, contract tiers on contacts, business impact on cases. They feed automations, analytics, and the Agent Stack.

Provision your tenant's BigQuery dataset, download a service account key, and point a BI tool at it.

Atender's Data Warehouse exposes your tenant's support data through a dedicated Google BigQuery dataset of pre-built SQL views — connect Looker, Tableau, Metabase, or query it directly.

Every template category, where it gets used, and what it gates the template to. Picking right at creation matters because the category controls which product surfaces can pick this template up.

Every structure and content block available in the Email Studio editor, what each does, and when to reach for it. Plus the per-block properties that work across all blocks.

A complete worked confirmation email template — block layout, copy, merge tags, and the automation rule that sends it. Copy-paste ready for any team that wants to acknowledge inbound emails automatically.

Insert dynamic placeholders into Text and Heading blocks so the email reads like it was written for the recipient. Covers contact fields, conversation fields, custom fields, and what happens when a merge tag can't resolve.

Wire an Email Studio template into an Automation rule using the Send Email Studio template action. Cover language strategy, merge tag context, and the X-Email-Template header for delivery diagnostics.

Build an email template from scratch — pick a category, drop in blocks, configure styling, and ship. Walks through a Confirmation Email example.

Add a language version of an existing template — translate the subject and body, keep the layout. Customers receiving emails in their own language is one of the easiest support quality wins available.

Five common rendering problems and what causes them: layout broken in specific clients, missing images, merge tags showing as literal text, dark-mode legibility, mobile breakage.

Email Studio is a visual drag-and-drop builder for the email templates Atender sends — automation replies, CSAT surveys, system notifications, user invitations. Design once, send everywhere, in any language.

Build a form from scratch or a template, configure fields, pick what each submission does, brand it, then share the URL or embed snippet.

Build branded, multilingual forms hosted by Atender — each submission can create a conversation, a CRM contact, a callback request, or a log entry.

Define the start trigger, ordered milestones, end conditions, and optional timeout for a new journey.

Journey Tracker defines named, multi-step flows that conversations pass through — start trigger, ordered milestones, end conditions — so you can measure where conversations slow down or stall.

Add your company logo so it shows up in the workspace switcher and as the browser favicon.

Atender auto-derives dark-mode variants from your brand colors. If the auto pick doesn't match your brand, override the dark variants manually.

Set the primary and optional accent colors that style buttons, tabs, highlights, and accents across the Atender agent UI.

The Look & Feel logo styles the Atender agent UI. Customer-facing surfaces — KB portal, Web Chat, Incidents, CSAT, emails — each have their own logo configuration.

Look & Feel controls your workspace logo and brand colors for the Atender agent UI — separately from the customer-facing surfaces, which each have their own branding.

When multiple opening-hours rules might apply to the same conversation, how Atender picks one. Plus the matching matrix (team + channel beats team beats channel) and what happens when no rule matches.

Two patterns for non-trivial schedules: split shifts (lunch break in the middle of the day) and multi-timezone teams (NYC + London + Singapore offices, each open in their own time).

Build your first opening-hours rule from scratch — name it, scope it, define a weekly schedule, set the timezone. Walks through a Standard Support example covering email and chat 9am-5pm weekdays.

Pick a country to import the public holiday calendar automatically, then add date overrides for company-specific closures (or partial-day hours like Christmas Eve 8-noon). Holidays default to closed; you can override.

Five usual suspects when a rule isn't behaving as expected: scope mismatch, active toggle off, another rule winning specificity, timezone mismatch, or a date override silently overriding.

Opening Hours defines when your support team is available — per team, per channel, per schedule, with timezone awareness and holiday calendars. Multiple independent rules means you can model complex coverage without contortions.

Update your name and email, choose icon colors, pick what happens after marking a conversation done, and set your default workspace.

Walk through the per-channel configuration — threshold, triggers, prompt message, timeout action — for Phone, Chat, SMS, and Email.

If a returning customer is starting fresh threads instead of being recognized, walk these five checks. The most common cause is a mismatched identifier.

When a customer who already has an open or recent conversation reaches out again, Atender asks whether it's the same issue or new — and either appends to their thread or starts a new one.

What each of the 5 built-in roles can do — Owner (everything, immutable), Team Lead (supervisor), Agent (frontline), Reader (read-only), Analytics (reporting). Plus when to use each as a starting point for a custom role.

The 8 permission categories — Conversations, Notes, Contacts, Teams, Users, Analytics, Settings, Modules — and the kinds of toggles each contains. Use this as a map when designing custom roles.

Build a role with exactly the permissions you need — name it, pick permissions per category, save. Covers starting from scratch and starting from a standard role.

Two paths — assign the role at user-creation time (during invite), or change an existing user's role later. Users can have multiple roles; their effective permissions are the union.

Five usual suspects when a user reports missing access: their role doesn't have the permission, they have multiple roles where one's restricting, the change hasn't propagated yet, they're on the wrong tenant, or the feature is feature-flagged off.

Roles bundle permissions; users get assigned roles. Atender ships 5 standard roles (Owner, Team Lead, Agent, Reader, Analytics) and lets you create custom roles with granular control over 60+ permissions across 8 categories.

The four SLA risk states, how a conversation moves between them, and what happens to the timer when a conversation reopens or closes. Includes the one-way escalation rule and the First Reply Time reopen exception.

The four priority bands — low, normal, high, urgent — are not agent-assigned priorities. They're time tripwires the timer crosses as it counts down. Understanding this is the difference between using SLAs effectively and arguing with the badge.

First Reply Time, All Messages, and Resolution Time — what each measures, when each starts and stops, and how to combine them in a single policy.

Complete worked example: a Standard Support SLA covering email and chat, with sane targets, the at-risk escalation automation, and a daily breach-rate review surface. Copy-paste values included.

Watch breach rate, average compliance, and per-policy performance over time. Where to find the dashboards and what to look for when the numbers move.

Give Premium / VIP customers tighter SLAs than standard ones, automatically. Today this works through team-based assignments paired with tier-based routing automation; tier→policy linkage in customer tier settings is partial.

Toggle a policy to ignore office hours so the timer counts continuously, regardless of when the conversation arrives. Use for after-hours teams, monitored channels, and contractual round-the-clock SLAs.

Build a working SLA policy from scratch — name it, pick metrics, set priority targets, attach it to channels and teams. Walks through a Standard Support example you can copy.

Use the assignment matrix (team + channel scoring) to give each channel its own response targets — chat gets minutes, email gets hours, voice routes to a different policy entirely. Without duplicating the rest of your SLA setup.

Three usual suspects when an SLA breaches overnight on a policy you thought respected office hours: the policy has Ignore office hours on, the workspace's opening hours are misconfigured, or a different policy is winning the assignment race.

No. Once a metric breaches, the historical record is permanent — by design. The right reaction is fast response and post-incident learning. Three things you can do that aren't unbreaching.

Service Level Agreements track response and resolution time commitments visually and in real time. Conversations climb through four risk states — SLA, At risk, Urgent, Breach — driving urgency, visibility, and automation hooks for escalation.

Authorize Slack, pick a notification channel, and choose which events trigger messages.

Connect a Slack workspace to Atender to receive notifications about assignments, customer replies, SLA breaches, and new conversations in a Slack channel.

Every dynamic variable available inside a snippet, what each resolves to, and what happens when the source data is missing. Plus the patterns that prevent awkward empty values.

Three ways to insert a snippet into a reply — shortcode, picker, favorites. Plus when each is fastest and the keyboard shortcuts that make snippet usage feel native to the reply flow.

Build a new snippet — pick a title, shortcode, content with variables, scope, and category. Personal scope is always available; Global is admin-only and can be restricted to specific teams.

Snippets are pre-written reply templates with dynamic variables. Type a shortcode like /refund and the snippet expands inline — fast, consistent, personalized. Saves agents the most repetitive part of replying.

When to make a tag a Topic (red — what the conversation is about) versus Metadata (blue — what's notable about it). Picking right keeps your taxonomy scannable and your filters useful.

A complete worked example: build a Returns parent tag with four sub-tags, write descriptions that auto-tag well, and wire one of them to an automation. Copy-paste values included.

Move tags under parents, re-parent existing tags, and rename them in place. The taxonomy is a tree — keep it shallow, named consistently, and reflective of how your team actually thinks.

Turn on the AI Auto-Tag switch so Sidekick applies a tag to conversations whose content matches the tag's description. The description is the only signal — write it carefully.

Add a new tag to your taxonomy — single tag or several at once via Quick Add. Pick a name, an optional description, the type (Topic or Metadata), and a parent.

Apply a tag to many conversations at once — filter the Monitor to the set you want, multi-select, and tag them in a single operation. Useful for retroactive cleanup after introducing a new tag.

Atender shows usage counts before deletion so you know what would be affected. The right move is usually to rename or re-parent the tag rather than delete it — deletion strips the tag from every conversation that has it.

Tags are the classification system that runs through all of Atender. They organize conversations, power filtering and views, drive automations, and help AI agents and routing decide where a conversation belongs.

A copy-paste pattern for organizing support into Tier 1 frontline, Tier 2 senior, and a specialist team. Uses team defaults on channels plus automation rules — Atender has no separate 'escalation' primitive; tiered support is just teams + reassignment.

Wire a channel — email inbox, voice number, web chat widget, etc. — to a default team so inbound conversations land with the right group. The default-team setting lives on the channel, not on the team.

Tie a team to a specific opening-hours rule so SLA timers, IVR routing, and chat availability all reflect the team's actual schedule. Most useful for multi-region teams or specialist teams with hours that differ from the workspace default.

Build a new team — pick a name, add initial members, optionally link an opening hours rule. The team becomes available for routing automations and SLA scoping immediately.

Update team membership — add new users, remove people who left, move someone between teams. Multi-team membership is supported; users keep their role across all their teams.

Teams group your agents into functional units — by skill, language, channel, region. Used by routing automations, SLA assignment matrix, opening hours rules, and analytics. Users can belong to multiple teams; no team hierarchy today.

Add product names, industry terms, and people's names to your custom dictionary so the transcription engine spells them correctly.

Open the feedback queue, filter to what needs attention, read details, comment, and move entries through the status flow.

A built-in collector for bug reports and feature requests from your team — submissions land in a triage queue with status, priority, module context, and comment threads.

Promote or demote a user — adjust permissions when responsibilities change, or when a teammate steps into supervisor or admin duties. Role changes apply immediately; in-flight conversations are not reassigned.

When a user is deactivated, their conversations, drafts, comments, and team memberships are preserved — nothing is auto-reassigned. Plan the handover before deactivating; the system protects history, not workflow continuity.

When an invite expires or doesn't arrive, resend a fresh one from the Pending Invites section. Old links stop working when a new invite is issued.

Send a new user an email invite that lets them set their password and join the workspace. Includes role assignment, optional team membership, and what the invitee experiences.

Deactivate a user when they leave the team or shouldn't have temporary access — they can't log in, but their conversation history is preserved. Reactivate later by toggling. Permanent deletion is a separate, irreversible action.

Diagnostic flow for the most common 'I can't log in' or 'I'm missing access' reports. Six things to check, in order, from authentication through permissions.

Users are the people with access to your Atender workspace — agents, supervisors, admins. Each gets a login, a role, optional team membership. Invitation-based onboarding via email; activation toggles preserve history without permanent deletion.