Limits

Every enforced size, count, timeout and cap on the platform. Where a field has a default and a range, both are given. None of these is configurable per workspace.

Agents

Limit	Value
`name`	At least 1 character, unique in the workspace. No maximum
`prompt`, `greeting`	No length limit
Tools, MCP servers, handoff targets per agent	No limit
`faqs` per agent or task	10
`max_steps` (LLM → tools rounds per turn)	1–50; `null` = 4 on voice and video, 25 on text
`max_duration_seconds`	30–10,800, default 1,800 (30 min); `null` = the 3-hour platform cap
`silence.timeout`	5–600 s, default 10
`silence.max_check_ins`	0–5, default 2
`keypad_input.timeout`	0–10 s, default 2
Voicemail message	1–2,000 characters
`tts.speed`	0.5–2.0, default 1.0
ElevenLabs `stability`, `similarity_boost`	0.0–1.0
Noise cancellation `enhancement_level`	0.0–1.0, default 0.8
`min_silence_duration`	At least 0.25 s, default 0.5
`handoffs[].recent_turns`	1–10, with `context: "summary"` only
`conversation.summary_limit`	At least 1
Analysis fields	25; names match `^[a-z][a-z0-9_]{0,39}$`
Members in an agent team	10
Handoffs executed in one call	25
Builtin web search options	xAI: 5 domains, 20 X handles. OpenAI: 100 allowed domains

There is no size limit on an agent config or on an inline definition sent with a call request.

FAQs

Limit	Value
FAQ `name`	1–80 characters, unique in the workspace (case-insensitive)
`question`	1–300 characters; duplicates in one FAQ refused (case-insensitive)
`answer`	1–2,000 characters
Entries per FAQ	500
Entries per `POST /v1/faqs/{faq_id}/entries`	1–500, all or nothing

Tools

Limit	Default	Range
`http` `timeout`	20 s	up to 120 s
`http` response body read	—	1 MB
`code` `timeout`	10 s	up to 30 s
`frontend_rpc` `timeout`	5 s	up to 60 s
`transfer` `ringing_timeout`	30 s	5–120 s
`transfer` DTMF digits	—	64
A test run, whole tool	—	60 s

Code sandbox

Limit	Value
Memory	128 MB
`console.log` output kept	100 lines, each cut at 2,000 characters
Input payload	2 MB
`fetch` from inside the sandbox	1 MB response, 10 s timeout, redirects not followed

Agent tasks

Limit	Default	Range
`max_steps`	25	up to 50
`timeout_seconds`	300	10–600
`output` fields	—	1–25; names match `^[a-z][a-z0-9_]{0,39}$`
`vars` (inputs)	—	No limit; names match `^[A-Za-z_][A-Za-z0-9_]*$`
Trace steps stored	—	200
One trace field	—	4,000 characters, then cut

See task errors and limits.

Calls and chats

Limit	Value
Call duration	3 hours on every voice and video call (phone, web, media stream), lower if the agent’s `max_duration_seconds` says so (default 30 min)
Outbound ring time	60 s before an unanswered call is given up
Warm-transfer briefing	90 s
Post-call analysis	Skipped when the caller spoke fewer than 2 turns
Chat inactivity	None. A chat stays open until `end_call` or `POST /v1/chats/{chat_id}/end`
Chat message body	100,000 characters
Chat message wait	120 s, then the response returns `status: "running"`. `stream: true` has no cap
Chat token lifetime	1 hour
`contact_key`	1–256 characters

Files

A file over a limit marked stored is kept for you, and the agent is told why it cannot read it. Nothing is truncated.

Limit	Value	Over it
Files per message	4	The rest are not received
One file, any type	20 MB decoded	Not received
Photo the agent sees	JPEG, PNG or WebP up to 10 MB, read from the bytes	Stored as a plain file
Stored photo resolution	Longest edge downscaled to 1,568 px
PDF pages	100	Stored, not read
PDF scanned pages	10	Stored, not read
Text in one document	100,000 characters	Stored, not read
Word file unzipped	50 MB, 1,000 parts	Stored, not read
`filename`	255 characters
Signed file URL, and the link a tool’s File parameter receives	Valid 1 hour

See vision and images.

Batches

Limit	Call batch	Email batch	WhatsApp batch
Recipients per create or add request	10,000	10,000	10,000
Recipients per batch	No limit	No limit	No limit
`name`	1–200 characters	1–200 characters	1–200 characters
Concurrency	`max_concurrency` 1–10, default 1	`draft_concurrency` 1–10, default 5	—
Gap	`dial_gap_seconds` 0–3,600, default 0	`send_gap_seconds` 1–3,600, default 1; `draft_gap_seconds` 0–3,600, default 0	`send_gap_seconds` 1–3,600, default 1
Daily cap (per local day)	`dial_daily_cap`, default `null`	`send_daily_cap`, default fixed 200	`send_daily_cap`, default fixed 250
Attempts	`max_attempts` 1–5, default 1	`draft_attempts` 1–5, default 1; `send_attempts` 1–5, default 3	`delivery_attempts` 1–5, default 3
Retry wait	`retry_after_minutes` 5–1,440, default 30	—	`delivery_retry_after_hours` 24–168, default 48
`start_at`	Up to 90 days ahead	Up to 90 days ahead	Up to 90 days ahead

Daily cap shape	Rule
Fixed	`limit` ≥ 1
Ramp	`start` ≥ 1, `end` > `start`, `step` ≥ 1, `interval_days` 1–30 sending days

Email from_name is at most 200 characters; one email send request names up to 10,000 recipients. An analysis backfill takes up to 100 call ids.

Observability and webhooks

Limit	Value
Observability date range	7 days; wider is `400 date range cannot exceed 7 days`. Call and conversation lists are not limited
Series in a breakdown before the rest fold into “Other”	8
Webhook delivery timeout	5 s
Webhook retries	None. One attempt per event
Webhook delivery log page	Default 50, max 200

Names and credentials

Limit	Value
Secret name	`^[A-Za-z_][A-Za-z0-9_]*$`, no length limit; no limit on count
Secret value, provider key	1–16,384 characters
Workspace name, personal access token name	1–100 characters
Carrier account name	1–120 characters
MCP tool name	200 characters
Integration `allowed_tools`	`null` (all) or at least 1 entry
`retention_days`	1–3,650, or `null` (forever, the default)
Dashboard session	30 days

Pagination

List endpoints take limit and offset and return { items, has_more, limit, offset }, with no cursor and no total. See pagination.

`limit` max, default	Endpoints
200, 200	Agents, tools, tasks, FAQs, secrets, calls, webhooks, credit ledger, telephony accounts, phone numbers, remote numbers, telephony providers, integrations, integration catalog, triggers, stream connections
200, 100	Conversations, conversation sessions, chats, call batch recipients
200, 50	Task runs, call batches, email batches, email batch recipients, email sends, WhatsApp batches, WhatsApp batch recipients, webhook deliveries
500, 200	Conversation items, chat items
100, 100	Catalog avatars

Two catalog endpoints differ: GET /v1/catalog/voices pages with page and returns total_count, and model search (GET /v1/catalog/models) takes limit up to 50 (default 20) with a cursor.

Limits that do not exist

No API rate limit. No per-token, per-workspace or per-IP quota and no 429 from Talqing. A 429 you see was relayed from one of your providers.
No workspace concurrency cap. Only a batch’s own max_concurrency or draft_concurrency limits parallelism, per batch.
No spend cap. The only stop is a credit balance of zero; low_balance_threshold ($2) only warns. See pricing and credits.

Get started

Build

Guides

Agents

FAQs

Tools

Code sandbox

Agent tasks

Calls and chats

Files

Batches

Observability and webhooks

Names and credentials

Limits that do not exist

​Agents

​FAQs

​Tools

​Code sandbox

​Agent tasks

​Calls and chats

​Files

​Batches

​Observability and webhooks

​Names and credentials

​Pagination

​Limits that do not exist

Agents

FAQs

Tools

Code sandbox

Agent tasks

Calls and chats

Files

Batches

Observability and webhooks

Names and credentials

Pagination

Limits that do not exist