Call History
The Call History section provides a comprehensive view of all calls made by your AI agents. You can search, filter, and analyze call data to monitor performance and identify insights with powerful filtering capabilities, an audit/review workflow, and customizable views.A banner now sits at the very top of this page, letting you switch at any time between the classic Call History table and the new, improved Calls view. Both views are documented separately — this page covers the improved view; head over to Call History (Classic) for a full walkthrough of the original table if you’re using that view instead.
Accessing Call History
Navigate to Calls in the main menu to access your complete call history. This now opens the improved calls view (Calls v2) by default.Calls v2 brings fully configurable, drag-and-drop columns (including dynamic input/output fields grouped by agent, with key columns pinned to the right edge), saved views, real AI latency metrics with a per-turn detail chart, an expanded filter bar (date presets, multi-agent selection, tag autocomplete, exact Call ID, latency range, and a one-click “suspicious calls” filter), a built-in review/audit workflow with per-call notes, and an Excel export that can target either all filtered calls or just your current selection.
Switching Between Views
A notice banner sits at the very top of the page reminding you that you’re on the new Calls view.1
Notice the Banner
At the top of the page, a highlighted banner reads “You’re viewing the new Calls page.”
2
Switch Back if Needed
Click Switch back to the classic view in the banner to be taken to
/calls-old, the original Call History table, if you run into something that doesn’t yet behave as expected in the new view.3
Return Anytime
Both views read from the same underlying call data, so you can move back and forth between them without losing information — only view-specific preferences like column layout or saved views live separately per view.
This same toggle works in both directions: whether you land here from the main menu or arrive via a direct link, the banner at the top of the page always offers a one-click way back to the classic table, and the classic table offers an equivalent way forward into this improved view. See Call History (Classic) for the documentation covering that table in detail.
Interface Overview
The calls interface consists of:Configurable Columns
The improved calls view lets you fully control which columns appear, in what order, and even add columns sourced directly from each agent’s input parameters or structured output — no need to open a call individually to see a specific data point.
Saved column layouts live in the separate 👁️ Views menu next to the Columns button — see Saved Views below.
1
Open the Column Manager
Click the ⚙️ Columns button next to the Export to Excel button to open the column configuration drawer.
2
Reorder Columns
In the Columns section, drag any row using the ⠿ handle to reorder it. The table updates its column order live as you drag.
3
Show or Hide Columns
Use the checkbox next to each column name to toggle its visibility. Columns marked as non-toggleable (such as Actions) always remain visible.
4
Add a Dynamic Column from Input or Output
Click Input columns or Output columns, find the field you need under the relevant agent group (grouped and collapsible per agent), and click Add. The field appears immediately as a new column, placed just before the pinned columns, and is tagged with an input/output badge in the Columns list.
5
Remove a Dynamic Column
Dynamic columns display a small ✕ button next to their badge in the Columns list. Click ✕ to remove them from your view entirely.
6
Reset to Defaults
Click Reset to default columns at the bottom of the panel to discard all customizations and restore the original column set.
Pinned Columns
A small number of columns are pinned to the right edge of the table so they stay visible regardless of how many columns you add or how far you scroll horizontally.
Saved Views
Beyond your single auto-saved layout, the 👁️ Views menu (next to the Columns button in the header) lets you keep multiple named views — for example, one focused on latency troubleshooting and another focused on audit review — and switch between them instantly.
1
Configure a Layout
Show, hide, reorder, and add dynamic columns until the table matches the view you want to save.
2
Open the Views Menu
Click 👁️ Views in the header to open the saved views panel.
3
Name and Save
Type a name into the New view name… field and click Save current. The current column configuration is stored as a new saved view.
4
Apply a Saved View
Click on any saved view’s name in the list to instantly replace your current column configuration with that view’s layout.
5
Delete a Saved View
Click the ✕ next to a saved view to remove it. This only deletes the saved view itself — it doesn’t change your currently active columns.
Legacy Column Configuration Migration
If you previously customized columns in the classic Call History table, that layout isn’t lost — the improved view automatically imports it the first time you open it.
Default Column Set
Out of the box, the improved view includes:
Additional columns such as To, Max latency, Company, Email, Summary, Type, Hung up by, and Agent duration are available but hidden by default — enable them from the column manager as needed.
Review & Audit Workflow
Every call can now be marked as reviewed and annotated with a free-text note directly from the table — no need to leave Call History to track your QA process.
1
Open the Review Popover
Click the Reviewed cell for any call to open its inline note editor.
2
Mark as Reviewed
Toggle the call’s reviewed status to indicate it has been audited.
3
Write an Audit Note
Type your observations, findings, or QA comments directly into the note field. Your input is debounced and saved automatically a moment after you stop typing.
4
Close the Popover
Click anywhere outside the popover to close it. Only one review popover can be open at a time, so opening a new one automatically closes any other that was active.
5
Revisit Later
Reopen the Reviewed cell for any call at any time to read or edit the existing note — your review history for that call is preserved.
AI Latency Metrics
Every call now tracks how quickly your AI agent responded throughout the conversation, giving you a direct signal of conversational responsiveness independent of call outcome or duration.
1
Scan Averages at a Glance
Sort the table by Avg AI latency to quickly identify calls or agents with consistently slow responses.
2
Spot Outliers with Max Latency
Enable the Max latency column (hidden by default) to catch calls with an isolated but severe delay that the average might hide.
3
Read the Sparkline
The Latency (turns) column renders a lightweight line for each call, plotting latency per turn. A rising line toward the end of a call can indicate the agent slowing down as the conversation got more complex.
4
Open the Full Latency Detail Modal
Click into a call’s details to open the dedicated AI latency modal: a full, interactive chart plotting latency per turn, with exact millisecond values on hover and the timestamp of each turn along the X axis.
Latency Detail Modal
The latency detail modal turns the compact table sparkline into a full-size, turn-by-turn chart so you can pinpoint exactly where a call slowed down.
Recording Player with Waveform
The recording player has been upgraded from a simple audio control to a full waveform-based player, built from the actual decoded audio of the call — not a generic placeholder animation. It also includes a rich, per-turn conversational view when detailed provider data is available.
1
Open the Player
Click the Recording column for any call with an available recording.
2
Review the Waveform
Wait for the audio to load and decode — the waveform renders automatically once ready, reflecting the real amplitude of the call.
3
Seek and Adjust Speed
Click on any part of the waveform to jump there instantly, or use the speed buttons to review a section more quickly or more carefully.
4
Cross-Reference with the Transcript
Use the transcript panel next to the player to follow along with what’s being said as you scrub through the recording. Expand latency or tool badges on any turn for a deeper technical breakdown of that moment in the call.
Excel Export
Exporting to Excel respects exactly what you see: only the columns currently visible in your table — in the order you’ve arranged them — are included in the exported file, including any dynamic input/output columns you’ve added and the current review status/notes.
1
Configure Your View
Use the column manager to show, hide, and reorder columns until the table reflects exactly the data you want to export. Apply a saved view first if you have one ready for this purpose.
2
Apply Your Filters
Narrow down the calls you want using the filter bar, since the full export follows the currently filtered result set.
3
Choose Full or Selective Export
Click Export to Excel to export every filtered call, or select specific rows with the checkboxes and click Export selection (N) to export only those calls — useful when you’ve hand-picked a set of calls during a review session.
4
Watch the Export State
The button shows an “Exporting…” state while the file is being generated.
5
Open the File
The downloaded spreadsheet mirrors your visible columns and their order, including any dynamic agent input/output fields you’ve added and the Reviewed status/notes, making it easy to share exactly the view you were working with.
Advanced Filter System
Filter Visibility Configuration
Customize which filters are visible in your interface to streamline your workflow and focus on the most relevant filtering options.
1
Open Filter Settings
Click the ⚙️ settings icon at the right end of the Filter Bar to open the filter visibility panel.
2
Toggle Individual Filters
Use the checkboxes to show or hide each of the following filters independently:
- Goal — filter calls by the agent’s configured goal or objective
- Tags — filter by dynamic or custom tags
- Evaluations — filter by evaluation completion or outcome
- Call ID — the direct database lookup search field, matched on the exact ID
3
Apply and Close
Close the panel by clicking outside of it or pressing the settings icon again. Hidden filters are immediately removed from the Filter Bar, freeing up space for the ones you actually use.
4
Automatic Persistence
Your filter visibility preferences are saved automatically per user and restored the next time you open Call History — no need to reconfigure them on every visit.
By default, the Tags filter is visible in the Filter Bar out of the box, so you don’t need to open the settings panel to start filtering calls by tags. Goal, Evaluations, and Call ID remain part of the same visibility configuration and follow their own default state, which you can adjust at any time.
Date Range Filter
The date range filter now offers a much wider set of quick presets, from short rolling windows measured in minutes to full calendar months, alongside a fully custom range with time precision.
1
Open the Picker
Click the date range field in the Filter Bar to reveal the list of presets.
2
Pick a Preset
Click any preset to apply it instantly — the field updates to reflect your selection and the table refreshes.
3
Or Set a Custom Range
Scroll to the Custom range section at the bottom of the picker and set an exact From and To date and time using the datetime inputs.
4
Confirm
Click Done to close the picker once your custom range is set.
Use the datetime picker to set precise time ranges for your searches. The filter automatically adjusts to your local timezone.
Multi-Select Agent Filter with Searchable Combobox
The Agent filter is a typeahead combobox that supports selecting multiple agents at once, so it stays fast and usable even when your account has hundreds of agents and you need to compare a handful of them together.
1
Start Typing
Click into the Agent field and type any part of the agent’s name. As soon as the field is focused, a dropdown of matching agents appears.
2
Filter as You Type
The list narrows in real time to agents whose name contains what you’ve typed, capped to the closest matches so the dropdown stays responsive.
3
Select One or More Agents
Click an agent from the list to add it as a chip inside the field. Keep typing and selecting to add additional agents — each selected agent appears as its own removable chip.
4
Remove a Single Agent
Click the ✕ on any chip to remove just that agent from the selection, keeping the rest of your filter intact.
5
Clear All Agents
Click the ✕ at the end of the field to remove every selected agent at once and return to viewing calls from all agents.
Status Filter
Filter by call status with visual indicators:Direction Filter
- Inbound - Calls received by your agents with 📞 icon
- Outbound - Calls initiated by your agents with 📲 icon
- All - Both directions
Success Filter
Filter by call outcome with enhanced evaluation criteria:- Successful - Calls marked as successful based on your evaluations
- Unsuccessful - Calls that didn’t meet success criteria
- All - All calls regardless of outcome
Goal Filter
The Goal filter lets you narrow down calls by the objective configured on the agent, making it easier to review calls tied to a specific business intent (e.g., booking, qualification, support resolution).
Latency Range Filter
Filter calls by their average AI response latency to isolate calls that were noticeably fast or slow, without needing to open the latency detail modal one call at a time.
1
Set a Minimum
Enter a value in Min latency to exclude calls that responded faster than that threshold.
2
Set a Maximum
Enter a value in Max latency to exclude calls that responded slower than that threshold.
3
Combine Both
Use both fields together to isolate a specific latency band, for example calls between 1500 ms and 3000 ms, to investigate a particular tier of responsiveness.
4
Clear the Range
Empty either field to remove that bound and widen the results again.
Suspicious Calls Filter
The “Suspicious calls” filter is a one-click way to surface calls that are more likely to need a closer look, without manually combining several filters yourself.
1
Enable the Filter
Toggle Suspicious calls on in the Filter Bar.
2
Review the Narrowed Results
The table updates to show only calls matching at least one of the suspicious conditions above.
3
Investigate Each Call
Open each flagged call’s details, recording, or latency chart to determine whether it needs a review note, a tag, or escalation.
4
Mark as Reviewed
Use the Reviewed column to mark calls you’ve checked and leave an audit note, so the next pass through your suspicious calls doesn’t re-cover the same ground.
Unlike Min/Max latency or Status, which apply a strict AND to your other filters, “Suspicious calls” applies an OR across its own internal conditions — a call needs to match only one of them to show up, since each condition on its own is a reasonable independent signal of a problematic call.
Exact Call ID Search
The Call ID filter performs a direct, exact-match database lookup and has exclusive priority over every other filter in the Filter Bar. As soon as you start typing a value into the Call ID field, all remaining filters (Date Range, Agents, Status, Direction, Success, Goal, Tags, Evaluations, Latency range, and Suspicious calls) become visually disabled and are excluded from the query — the interface greys them out and blocks interaction with them to make it clear that only the Call ID lookup is active.Think of it as a small measurement problem: the moment you observe the system through a Call ID, every other possible filter state “collapses” — much like a quantum measurement collapsing a wave function into a single outcome. Before you type an ID, the Filter Bar exists in a superposition of Date Range, Agents, Status, Direction, Success, Goal, Tags, Evaluations, Latency range, and Suspicious calls, all potentially active at once. The instant a Call ID value is present, that superposition resolves into a single, deterministic state: exactly one call, found by its exact ID, with no other filter contributing to the result.
1
Enter Call ID
Type the specific call ID in the dedicated search field. You can also copy an ID directly from the Call ID column’s compact badge (or the row’s Copy ID quick action) and paste it here. As soon as a value is present, the rest of the filters switch to a disabled visual state.
2
Database Query
The system performs a direct database lookup for the exact call ID, ignoring any values currently set in other filters. Partial or fuzzy matches are not returned — the ID must match exactly.
3
Instant Results
The matching call is displayed immediately, bypassing all other filters.
4
Clear Search
Remove the call ID to restore normal filtering functionality. All previously configured filters are automatically re-enabled and reapplied — the “superposition” is restored exactly as you left it.
Dynamic Tags Filter with Autocomplete
Filter calls by dynamic tags that are automatically generated based on call content, outcomes, and agent behaviors, using a chip-based multi-select field with live autocomplete.
1
Start Typing a Tag
Click into the Tags field and start typing. A dropdown appears listing existing tags that match what you’ve typed.
2
Select a Suggested Tag
Click a suggestion to add it as a chip in the field. You can add multiple tags this way to filter for calls matching any of them.
3
Remove a Tag
Click the ✕ on any tag chip to remove just that tag, or clear the field entirely to remove all tag filters at once.
- Dynamic Tag Generation - Tags are automatically created based on call analysis
- Autocomplete Suggestions - Existing tags are suggested as you type, so you don’t need to remember exact spelling
- Multi-Select Capability - Select multiple tags to find calls matching any or all criteria
- Real-time Updates - Tag list updates as new calls are processed
- Custom Tag Support - Manual tags can be added to calls for additional categorization
The Tags filter is visible by default in the Filter Bar, so it’s ready to use as soon as you open Call History. You can still hide it from the ⚙️ filter settings panel if you prefer a more minimal filter set. Like all other filters, it is temporarily disabled while a Call ID search is active.
Evaluations Filter
Filter calls based on evaluation status, letting you quickly locate calls that are pending review, partially evaluated, or fully completed.
- Completed - Calls with all evaluations processed
- Pending - Calls waiting for evaluation
- Partial - Calls with some but not all evaluations completed
Filter Presets
Create and manage filter presets:1
Configure Filters
Set up your desired combination of filters (date range, agents, status, latency range, suspicious calls, etc.)
2
Save as Preset
Click “Save Preset” and give it a descriptive name
3
Quick Access
Access saved presets from the preset dropdown for instant filter application
4
Manage Presets
Edit or delete presets as your filtering needs evolve
Enhanced Visual Indicators
Filter Status Indicators
The interface now includes improved visual feedback:Selection Status
When calls are selected:- Selection Counter - Shows “Export selection (N)” in the header once at least one call is checked
- Bulk Action Bar - Appears with available actions for selected calls
- Clear Selection - One-click option to deselect all calls
Call Table
The call table displays key information for each call with enhanced visual design:The LLM, LLM detail, Avg AI latency, Max latency, and Latency (turns) columns are optional. In the classic Call History table they’re hidden by default and can be enabled from the column customizer; in the improved calls view, Avg AI latency and Latency (turns) are visible out of the box.
LLM and LLM Detail Columns
These two columns surface the underlying language model activity for every call, which is especially useful for agents built on ElevenLabs’ conversational stack, where a model cascade can automatically fall back to a secondary model mid-call if the primary one becomes unavailable or degrades.
1
Enable the Columns
Open the column manager above the Call Table and toggle on LLM and/or LLM detail to add them to your view.
2
Read the LLM Column
Each row shows the model(s) used for that call. A single model name means the call ran entirely on the primary model with no fallback.
3
Spot Degradations
When a call shows more than one model in the LLM column, it means the ElevenLabs cascade degraded at least once during the call — falling back from the primary model to a secondary one. This is the same detection logic used in the call details modal.
4
Drill Into LLM Detail
Expand the LLM detail column (or open the call in the Details Modal) to see the exact timestamp of each turn and which model produced it, helping you pinpoint when and why a degradation happened.
Bulk Actions
- Selection
- Export
When you select multiple calls:
- The “Export selection (N)” button appears in the header, next to the general Export to Excel button
- Bulk export options become available immediately, no separate confirmation step required
- Additional bulk actions appear in the action bar
- Deselect individual rows or clear the whole selection to hide the selective export button again
Enhanced Sorting
Click column headers to sort with improved indicators:- Single Click - Ascending order with ↑ arrow
- Double Click - Descending order with ↓ arrow
- Sort Memory - Last sort preference is remembered
- Multi-Column Sort - Hold Shift to sort by multiple columns
Pagination
Enhanced pagination controls:- Smart Page Sizes - 25, 50, 100, 300 calls per page
- Jump to Page - Direct page number input
- Results Counter - “Showing X to Y of Z calls” with total count
- Performance Optimization - Efficient loading for large datasets