Skip to main content
API Reference

Bots

Create and manage AI communication agents, verify channel deployment, and load their knowledge base.

Overview

A bot is a configured AI agent bound to a channel. The type is fixed at creation and determines how the bot is reached:

typeChannel
whatsappWhatsApp text conversations
whatsapp_voiceWhatsApp voice notes
inbound_callPhone calls customers place to you
outbound_callCalls the bot places (see Voice for availability)
ivrSmart IVR menu flows

Credential class: dashboard JWT or a cm_ key with scopes

Both credentials work.

Authorization: Bearer <jwt_access_token>
# or
Authorization: Bearer cm_your_api_key
EndpointsScope needed by a cm_ key
List, get, get prompt, tool catalog, skill catalog, config schema, validate config, probe MCP servers, list/get versionsbots:read
Create, update, put prompt, patch config, delete, toggle, deploy verify, create/edit/commit/roll back versionsbots:write
Knowledge listknowledge:read
Knowledge add, upload, deleteknowledge:write

A dashboard JWT bypasses the scope check. One extra role rule applies to JWT callers: POST /{bot_id}/deploy/verify requires owner or admin.

Base URL for every example: https://api.callmissed.com. Errors are {"detail": "..."}; validation failures are 422. A missing scope returns 403 naming the scope.

Credential redaction in config

config is a free-form JSON object. On the way out, any key whose name looks like a credential (access_token, api_key, auth_token, app_secret, secret, password, token, verify_token, and similar, matched case-insensitively at any depth) is replaced with "[REDACTED]". The stored value is unchanged; you simply cannot read it back.

PATCH /{bot_id}/config goes further and refuses a credential-looking key outright with 422 Credentials cannot be stored in bot config. POST and PUT still accept one, so prefer PATCH when a machine is writing the config.

What is validated at write time

Most of config is stored as-is, which is why Build an agent programmatically exists: call GET /bots/config-schema to learn the keys, and POST /bots/validate-config to catch the rest before you write. These sub-keys are the exception — they are schema-checked on every write path (POST, PUT, PATCH /config) and return 422 when malformed:

Sub-keyChecked
analysis_variablesPost-call extraction schema. A number field may add "display": "stars" and "max": 5 (default) or 10 to treat it as a 1 to max rating, shown as stars in the console; a value outside that scale is stored as the field's default (or null)
input_variablesPer-call {{token}} declarations
stt_keytermsList of at most 100 unique strings, each 1-100 chars. Applied by the Deepgram Nova-3 and Flux models, the AssemblyAI models, microsoft-speech-stt and gnani-prisma-v3.0; other speech models ignore it
pronunciationsList of at most 100 { "word", "say_as" } objects, each 1-64 chars, unique word. The voice says say_as instead of word (whole-word, case-insensitive); transcripts keep the original text
web_widget_enabledBoolean
web_widgetObject; title ≤ 100 chars, greeting ≤ 500 chars, color a hex colour
fallback_messageString ≤ 4000 chars
mcp_serversList of at most 10 objects, each with a string id and an https:// url
commandsList of at most 30 {name, instruction}; names are lower-cased, must be unique, 1-32 chars of letters, digits, spaces, - or _; instructions 1-2000 chars
disabled_toolsList of at most 50 tool names, each 1-100 chars, de-duplicated
button_actionsList of at most 20 rules for WhatsApp button taps (WhatsApp button actions); button 1-25 chars and unique ignoring case, action one of ai_reply, send_text, send_document, send_template, plus that action's fields
callback_callsObject (WhatsApp callback calls); enabled and allow_other_number booleans, voice_bot_id (required when enabled) and from_number_id ids of your own agent and active phone number, hours_start 0-23, hours_end 1-24 and later than hours_start, max_per_day 1-10. Stored with every field filled in
wa_details_template, wa_reengage_templateName of one of your approved WhatsApp templates (Templates outside the 24-hour window), with an optional wa_details_template_language / wa_reengage_template_language. Checked when you change it; null or "" clears it and its language
followup_agentObject; enabled and send_recap booleans, instructions a string ≤ 4000 chars, max_attempts an integer 1-5
call_personal_whatsappObject (Call personal WhatsApp); enabled, a boolean, and session_id, a linked number's id or null. Stored as { "enabled": …, "session_id": … }. Turning it on or changing session_id is owner-only (403 otherwise)
transcript_retention_days, recording_retention_daysWhole number of days, 1-3650. Applies to calls made after you save this; older calls follow your account's retention setting. The server records when each was saved as transcript_retention_set_at / recording_retention_set_at (Voice data retention)
trust_caller_idBoolean. When true, an inbound phone caller is identified by their caller ID (Who the caller is)
data_retention_mode"standard" or "zdr" (Voice data retention)
tts_fallbacksList of at most 3 { "tts_model", "voice" } objects (voice optional). Each tts_model must be a voice model that speaks the agent's language, each voice one of that model's voices; no repeats (Backup voices)

Two normalisations happen silently on every write: commands and disabled_tools come back normalised (lower-cased, de-duplicated) rather than as you sent them, and the retired voice_fallbacks key is dropped. Everything else in config is stored verbatim.

WhatsApp button actions

config.button_actions decides what happens when a customer taps a quick-reply button (for example on a template you sent them) on a WhatsApp Business number linked to this agent. The action runs as configured, without the AI choosing it.

{
  "button_actions": [
    {"button": "Send itinerary", "action": "send_document", "document": "itinerary", "text": "Here is the full itinerary"},
    {"button": "Price and inclusions", "action": "send_text", "text": "The 5-day trip is Rs 24,999 per person, stay and breakfast included."},
    {"button": "Talk to the team", "action": "send_template", "template": "callback_confirm", "language": "en", "then_ai_reply": true},
    {"button": "Something else", "action": "ai_reply"}
  ]
}
FieldRequiredNotes
buttonYesThe button's text, 1-25 chars (WhatsApp's quick-reply limit). Matched against the tapped button's text or payload, ignoring case and extra spaces
actionYesai_reply, send_text, send_document or send_template
textsend_text: yes (≤ 4096 chars). send_document: optional caption (≤ 1024 chars)
documentsend_documentThe name of one of the agent's documents
templatesend_templateAn approved template on the same WhatsApp account as the number. Only templates with no variables and no media header can be sent this way
languageNo (send_template)The template's language code. Needed only when the template is approved in more than one language
then_ai_replyNotrue lets the agent also reply after the action. Default false. Not allowed on ai_reply

A tap with no matching rule, or an ai_reply rule, is answered by the agent like any other message. If an action cannot be sent (the document or template is missing, a template needs values, or WhatsApp's 24-hour window has closed), the agent answers instead.

WhatsApp callback calls

config.callback_calls lets a WhatsApp agent arrange a phone call when a customer asks for one ("can someone call me at 5pm?"). While enabled is true the agent gets the request_callback tool. It queues a call to the customer's WhatsApp number, right away or at the time they ask for, placed by the voice agent you name from one of your phone numbers. That voice agent opens by saying why it is calling and is briefed with the customer's reason and the last messages of the chat.

{
  "callback_calls": {
    "enabled": true,
    "voice_bot_id": "c2d3e4f5-6789-4abc-8def-0123456789ab",
    "from_number_id": null,
    "hours_start": 9,
    "hours_end": 21,
    "max_per_day": 3,
    "allow_other_number": false
  }
}
FieldDefaultNotes
enabledfalseOffers request_callback to the agent on WhatsApp chats. It is not listed in GET /bots/tool-catalog and is never offered on a call
voice_bot_id—The agent that places the call. Required when enabled
from_number_idnullOne of your active phone numbers. null uses the number whose outbound agent is voice_bot_id, else your oldest active number
hours_start, hours_end9, 21Calls are placed only from hours_start:00 up to hours_end:00 in your workspace timezone. Asked for outside them, the agent asks the customer for another time
max_per_day3Calls back per customer in any 24 hours
allow_other_numberfalseWhen true, the customer may ask to be called on another number in the same country as their WhatsApp number

What the agent checks before it promises a call: the number is not on your do-not-call list for calls, you have credit, no call to that number is already on the line, the time is within 7 days and inside the hours above, and the daily limit is not reached. A customer has at most one call waiting per agent: asking again with a new time moves it. The call is placed within about a minute of its time, with the same do-not-call, credit and destination checks as any other call. If it cannot start because all your lines are in use or the phone network errors, it is tried up to twice more; a customer who does not answer is not called again. See and cancel waiting calls with the callback endpoints.

WhatsApp templates outside the 24-hour window

WhatsApp delivers a plain message only within 24 hours of the customer last messaging or calling your WhatsApp number; outside that window only an approved template is delivered. A phone call never opens the window, so when a caller asks your agent to "send me the details on WhatsApp", the text can only go out inside a template.

{
  "wa_details_template": "requested_details",
  "wa_details_template_language": "en",
  "wa_reengage_template": "reply_to_continue"
}
KeyNotes
wa_details_templateAn approved template with exactly one variable, in its body, and no media header or button variable. The text the agent sends fills that variable; with the template's own words it must fit WhatsApp's 1024-character body limit, and a longer text is refused so the agent can shorten it
wa_reengage_templateUsed when no details template is set: an approved template with no variable, or one body variable (filled with the customer's name). The details are not in it; the agent tells the customer they follow once they reply
wa_details_template_language, wa_reengage_template_languageThe template's language code. Needed only when the template is approved in more than one language

A write that sets or changes one returns 422 when the name is not one of your approved templates (in that language, when given), or when the template has a different shape. With neither set, the agent tells the customer it could not send the message on WhatsApp.

Call personal WhatsApp

call_personal_whatsapp: { "enabled": true } lets a calling agent message the caller on WhatsApp from your linked personal number during the call, with no template: whatsapp_personal_message_caller (a message, optionally with one file) and whatsapp_personal_remind_caller (one scheduled message for later). Both go only to the caller. Off by default, and nothing happens until a personal number is linked in the console. See Messaging the caller from your personal WhatsApp. A write with any other field, a non-boolean enabled or a session_id that is not an id returns 422. Turning it on, or changing session_id, needs the workspace owner's own console session: an API key, another team member or an agency stand-in gets 403, because the messages go from the owner's own phone. Turning it off is open to anyone who can edit the agent. With several personal numbers linked, session_id picks which one sends (null = the default); a named number that is not connected sends nothing. Linked personal numbers are listed only in the console. It is not copied when an agency provisions an agent into a client workspace.


GET/api/v1/bots

Lists your tenant's bots, newest first. No parameters.

curl https://api.callmissed.com/api/v1/bots \
  -H "Authorization: Bearer cm_your_api_key"
[
  {
    "id": "b1f2c3d4-5678-90ab-cdef-1234567890ab",
    "tenant_id": "a0b1c2d3-4455-6677-8899-aabbccddeeff",
    "name": "Support Bot",
    "type": "whatsapp",
    "system_prompt": "You are a friendly support agent for Acme.",
    "config": { "phone_number_id": "109876543210987", "access_token": "[REDACTED]" },
    "is_active": true,
    "created_at": "2026-06-06T12:00:00Z",
    "updated_at": "2026-06-06T12:00:00Z",
    "conversation_count": 42,
    "config_warnings": []
  }
]

403 when a cm_ key lacks bots:read.

GET/api/v1/bots/tool-catalog

Agent tools a bot can enable. Write the chosen name values into config.tools; the runtime resolves them from the registry. No parameters. Requires bots:read.

curl https://api.callmissed.com/api/v1/bots/tool-catalog \
  -H "Authorization: Bearer cm_your_api_key"
[
  {
    "name": "shopify_order_status",
    "label": "Look up a Shopify order",
    "summary": "Checks an order's payment, fulfilment and tracking by order number or email.",
    "description": "Look up a Shopify order's payment + fulfillment status and shipping tracking, by order number (e.g. 1001) or the customer's email.",
    "category": "general",
    "parameters": {
      "type": "object",
      "properties": {
        "order_number": { "type": "string", "description": "Order number, e.g. '1001' (with or without #)." },
        "email": { "type": "string", "description": "Customer email to find their recent orders." }
      }
    },
    "unavailable_on": []
  }
]

label and summary are a short name and a one-line plain-English description for showing the tool to people; description is the text the model reads when deciding to call it.

parameters is a JSON Schema object, empty ({"type": "object", "properties": {}}) for tools that take no arguments.

unavailable_on lists the channels a tool cannot run on, and is [] for most tools:

ValueMeaning
voiceThe call runtime never offers this tool on a phone or WhatsApp call. Enabling it on a calling agent does nothing; POST /api/v1/bots/validate-config returns an incompatible warning for it
whatsapp_personalNot offered on a linked personal WhatsApp account

Copy the name verbatim. A name that is not in this catalog is dropped silently when the agent starts — no error at write time, no error at call time, the tool simply is not there.

GET/api/v1/bots/skill-catalog

Skills a bot can enable. Write the chosen name values into config.skills. No parameters. Requires bots:read.

A skill is a bundle, not a flag: switching one on adds the agent tools it names and injects that skill's own instructions into the agent's prompt when the model loads it mid-conversation. So it changes how the agent behaves, not only what it can do — which is why the catalog publishes both halves.

Skills apply to chat channels only. A calling agent never reads config.skills: the key is resolved on the chat path, and nothing on a call looks at it. Put the tools a call needs directly in config.tools.

curl https://api.callmissed.com/api/v1/bots/skill-catalog \
  -H "Authorization: Bearer cm_your_api_key"
[
  {
    "name": "order_support",
    "description": "Look up Shopify orders, products, and customers, and start a draft order/checkout.",
    "tool_names": [
      "shopify_order_status",
      "shopify_product_lookup",
      "shopify_customer_lookup",
      "shopify_create_draft_order"
    ],
    "channels": ["text"],
    "unavailable_on": ["voice"],
    "instructions_preview": "Order-support rules: - Use shopify_order_status to check an order...",
    "instructions_truncated": true,
    "note": "Chat channels only. A calling agent never reads config['skills'], so enabling this skill on one does nothing — put the tools it bundles in config['tools'] instead."
  }
]
FieldMeaning
tool_namesThe agent tools this skill adds when it loads. Each one is also in GET /api/v1/bots/tool-catalog
channelsWhere the skill is resolved. Always ["text"]
unavailable_onSame vocabulary as the tool catalog. Always ["voice"]
instructions_previewThe start of the instruction text the skill injects, capped at 400 characters
instructions_truncatedtrue when the real instructions are longer than the preview

An unknown skill name in config.skills is dropped silently at conversation time, exactly like an unknown tool name. Setting skills on a calling agent returns an incompatible warning from POST /api/v1/bots/validate-config.

POST/api/v1/bots

Returns 201.

FieldTypeRequiredConstraints
namestringYes1-255 chars
typestringYesOne of whatsapp, inbound_call, outbound_call, ivr, whatsapp_voice. Immutable after creation
system_promptstringNoMax 20000 chars, default "". The voice runtime adds its own call-handling instructions on top
configobject | nullNoFree-form JSON. Only the sub-keys listed above are validated; everything else is stored as-is and reported in config_warnings
curl -X POST https://api.callmissed.com/api/v1/bots \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Support Bot",
    "type": "whatsapp",
    "system_prompt": "You are a friendly support agent for Acme. Keep replies under 3 sentences.",
    "config": { "tools": ["shopify_order_status"] }
  }'
{
  "id": "b1f2c3d4-5678-90ab-cdef-1234567890ab",
  "tenant_id": "a0b1c2d3-4455-6677-8899-aabbccddeeff",
  "name": "Support Bot",
  "type": "whatsapp",
  "system_prompt": "You are a friendly support agent for Acme. Keep replies under 3 sentences.",
  "config": { "tools": ["shopify_order_status"] },
  "is_active": true,
  "created_at": "2026-08-04T11:00:00Z",
  "updated_at": "2026-08-04T11:00:00Z",
  "conversation_count": 0,
  "config_warnings": []
}

403 without bots:write; 422 on an unknown type, a name outside 1-255 chars, a prompt over 20000 chars, or a malformed analysis_variables / input_variables block.

GET/api/v1/bots/{bot_id}

Same object shape as one row of the list, with an accurate conversation_count. Requires bots:read.

curl https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab \
  -H "Authorization: Bearer cm_your_api_key"

404 Bot not found when the id is not in your tenant; 422 when it is not a valid UUID.

PUT/api/v1/bots/{bot_id}

Partial update. Omitted fields are unchanged. type cannot be changed. Requires bots:write.

FieldTypeRequiredConstraints
namestring | nullNo1-255 chars
system_promptstring | nullNoMax 20000 chars. Written as one flat string — see the warning below
configobject | nullNoReplaces the whole object, not a deep merge

system_prompt here is the whole prompt as one block. A stored prompt is normally section-structured — the dashboard edits it as an objective, response guidelines and a conversation script — and writing a flat string through this route collapses those three boxes into one, which the owner then has to split apart by hand. Nothing reports it. Use PUT /api/v1/bots/{bot_id}/prompt, documented below, instead — unless you genuinely mean to replace the prompt with unstructured text.

curl -X PUT https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"system_prompt": "You are a concise support agent for Acme."}'

Returns the updated bot. Note that conversation_count is 0 in the update response; read the bot to get the real count.

403 without bots:write; 404 when not found; 422 on validation failure.

GET/api/v1/bots/{bot_id}/prompt

The agent's instructions split into the four boxes its owner edits in the dashboard. Requires bots:read.

system_prompt is stored as one string, but it is not written as one. The first three boxes are sections of that string, delimited by marker lines; the fourth, the first message, is config.greeting. This route is the published reader of that layout, so a program does not have to know the markers to edit one section.

curl https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/prompt \
  -H "Authorization: Bearer cm_your_api_key"
{
  "objective": "You are the voice of Acme. Understand what the caller needs before offering anything.",
  "response_guidelines": "- Friendly and consultative. Short sentences.\n- One question at a time.\n- Never guess a price or a date.",
  "conversation_script": "## 1. Open\n\nGreet the caller and ask how you can help.",
  "first_message": "Hi, thanks for calling Acme. How can I help?",
  "freeform": null
}

Exactly one side of the response is populated:

FieldMeaning
objectiveWho the agent is and what it is trying to achieve. null when the prompt has no sections
response_guidelinesHow it should speak. null when the prompt has no sections
conversation_scriptThe steps it works through, in order. null when the prompt has no sections
first_messageThe line it speaks when the call connects — config.greeting. null when there is none
freeformThe whole prompt, when it carries no section markers. null for a section-structured prompt

A prompt written before the section editor existed — or one written flat through PUT /api/v1/bots/{bot_id} — comes back in freeform with the three blocks null. That is the honest reading, not an empty agent: the reader never invents sections that are not there. Reading is deliberately strict, so a prompt with text above the first marker, or with the same marker twice, is also reported as freeform rather than guessed at.

404 Bot not found when the id is not in your tenant; 403 without bots:read.

PUT/api/v1/bots/{bot_id}/prompt

Writes the prompt back in the same four boxes, with the markers the dashboard parses. Requires bots:write.

FieldTypeRequiredConstraints
objectivestring | nullNoMax 8000 chars. The three blocks together must fit in 20000 chars, or the call returns 422
response_guidelinesstring | nullNoMax 8000 chars
conversation_scriptstring | nullNoMax 8000 chars
first_messagestring | nullNoMax 500 chars. Stored as config.greeting
freeformstring | nullNoMax 20000 chars — the same cap as system_prompt. The whole prompt as one block. Cannot be combined with the three fields above
curl -X PUT https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/prompt \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "objective": "You are the voice of Acme. Understand what the caller needs before offering anything.",
    "response_guidelines": "- Friendly and consultative. Short sentences.\n- One question at a time.",
    "conversation_script": "## 1. Open\n\nGreet the caller and ask how you can help.",
    "first_message": "Hi, thanks for calling Acme. How can I help?"
  }'

Returns the whole bot, the same shape PUT /api/v1/bots/{bot_id} returns, including config_warnings.

Three rules decide what a request changes:

  • Omitted means unchanged. Sending only first_message edits the greeting and leaves the prompt alone. Sending only prompt fields leaves the greeting alone.
  • Among the three blocks, omitted means empty. They are written together, so a block you leave out is cleared. GET first, change the one block you mean, and send all three back.
  • freeform and the blocks are mutually exclusive. They are two spellings of the same field; sending both is 422. Use freeform only for a prompt GET already returned as freeform — round-tripping one that way stores it byte-identically, with no markers injected.

first_message goes through the same merge as PATCH /api/v1/bots/{bot_id}/config, so the config validator still runs and the findings come back in config_warnings. An empty or whitespace-only first_message removes greeting rather than storing a blank one, which is the difference between "no greeting" and "a greeting that is nothing".

As with PUT /api/v1/bots/{bot_id}, rewriting the prompt on a non-WhatsApp-voice agent drops a leftover config.voice_system_prompt, so a retired spoken persona cannot keep answering calls.

StatusCause
403Key missing bots:write
404Bot not found
422freeform sent together with a section block, a field over its length limit, or a first_message the config validator rejects

PATCH/api/v1/bots/{bot_id}/config

Changes individual config keys without resending the whole object — the safe write for an agent that only knows about the keys it cares about. PUT replaces config wholesale and will silently drop every key you left out; this does not. Requires bots:write.

FieldTypeRequiredConstraints
valuesobjectNoKeys to set or overwrite. At most 50 keys, each 1-100 characters. Defaults to {}
unsetarray of stringNoKeys to remove. At most 50, each 1-100 characters. Defaults to []

The merge is: apply values over the stored config, drop every key named in unset, then validate the result. A key may not appear in both lists.

curl -X PATCH https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/config \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "values": { "greeting": "Namaste! Acme support, how can I help?", "max_call_duration_seconds": 600 },
    "unset": ["fallback_message"]
  }'

Returns the whole bot with the merged config:

{
  "id": "b1f2c3d4-5678-90ab-cdef-1234567890ab",
  "tenant_id": "a0b1c2d3-4455-6677-8899-aabbccddeeff",
  "name": "Support Bot",
  "type": "inbound_call",
  "system_prompt": "You are a friendly support agent for Acme.",
  "config": {
    "voice_model": "gpt-oss-120b",
    "tts_model": "deepgram-aura-2",
    "stt_model": "deepgram-flux-general-en",
    "voice": "Thalia",
    "language": "en",
    "greeting": "Namaste! Acme support, how can I help?",
    "max_call_duration_seconds": 600
  },
  "is_active": true,
  "created_at": "2026-08-04T11:00:00Z",
  "updated_at": "2026-08-04T11:31:12Z",
  "conversation_count": 0,
  "config_warnings": []
}

As with PUT, conversation_count is 0 in this response; read the bot to get the real count.

StatusCause
403Key missing bots:write
404Bot not found
422A config key cannot be set and removed together, Config keys must be unique, Credentials cannot be stored in bot config, or a malformed value for one of the validated sub-keys above

Removing a key is not the same as setting it to null: unset deletes it, and the runtime then falls back to that key's documented default. Setting it to null stores a null and is read back as one.

GET/api/v1/bots/config-schema

The machine-readable description of config: every key the runtime reads, its type, its default, what happens on a real call when you leave it out, and — where the allowlist is closed — the values it accepts. Fetch this before composing a config rather than guessing key names. Requires bots:read.

ParamTypeRequiredNotes
bot_typestringNoOne of whatsapp, inbound_call, outbound_call, ivr, whatsapp_voice. Narrows the result to the keys that apply to that channel, and builds minimal_example for it. Omit it to get every key. Any other value is 422
curl "https://api.callmissed.com/api/v1/bots/config-schema?bot_type=inbound_call" \
  -H "Authorization: Bearer cm_your_api_key"
{
  "bot_type": "inbound_call",
  "keys": [
    {
      "key": "tts_model",
      "type": "string",
      "channels": ["voice"],
      "summary": "The model that speaks the agent's replies.",
      "default": "bulbul:v3",
      "if_omitted": "Falls back to the Indic-tuned default voice, which sounds wrong on an English line.",
      "allowed_values": ["aura-2-en", "aura-2-es", "bulbul:v3", "bulbul:v4-flash", "deepgram-aura-1", "deepgram-aura-2", "deepgram-flux-tts", "fish-audio-drama-3", "gnani-timbre-v3.0", "gpt-4o-mini-tts", "gradium-tts-beta", "inworld-tts-2", "inworld-tts-2-flash", "mai-voice-2.1", "mai-voice-2.1-flash", "melotts", "microsoft-neural-tts", "rime-coda", "sonic-3.6"],
      "values_depend_on": null,
      "bounds": null,
      "validated_on_write": false
    },
    {
      "key": "voice",
      "type": "string",
      "channels": ["voice"],
      "summary": "Which speaker the chosen tts_model uses.",
      "default": null,
      "if_omitted": "The chosen tts_model's own default speaker is used.",
      "allowed_values": null,
      "values_depend_on": "tts_model",
      "bounds": null,
      "validated_on_write": false
    },
    {
      "key": "max_call_duration_seconds",
      "type": "integer",
      "channels": ["voice"],
      "summary": "Hard cap on one call, in seconds.",
      "default": 600,
      "if_omitted": "The call runs to the platform ceiling.",
      "allowed_values": null,
      "values_depend_on": null,
      "bounds": [30.0, 14400.0],
      "validated_on_write": false
    }
  ],
  "minimal_example": {
    "name": "Support line",
    "type": "inbound_call",
    "system_prompt": "## Objective\n\nYou are the voice of [your business]…",
    "config": {
      "voice_model": "gemma-4-31b",
      "tts_model": "deepgram-aura-2",
      "stt_model": "deepgram-flux-general-en",
      "voice": "Thalia",
      "language": "en",
      "max_call_duration_seconds": 600,
      "greeting": "Hi {{callee_name}}, thanks for calling. How can I help you today?"
    }
  },
  "notes": [
    "Every key is optional. minimal_example is the smallest body that creates an agent that answers a call properly.",
    "Nothing here is enforced on write: a config with errors is still saved, and the findings come back in config_warnings.",
    "voice and language are only valid in combination with tts_model — send the stack together and validate it before you save it.",
    "GET /api/v1/models lists every model id; the voice list for a speech model is on its own model entry.",
    "system_prompt is not one block: the dashboard edits it as an objective, response guidelines and a conversation script… use GET/PUT /api/v1/bots/{bot_id}/prompt instead of writing a flat prompt."
  ],
  "voice_tiers": [ { "id": "t1", "label": "Standard", "price_credits_per_minute": 4.0, "billing": "flat", "minimum_seconds": 30.0, "…": "…" } ],
  "stt_languages": { "deepgram-flux-general-en": ["en"], "…": ["…"] },
  "multilingual": { "stt": ["deepgram-flux-general-multi", "gnani-prisma-v3.0", "saaras:v4", "…"], "tts": ["bulbul:v3", "sonic-3.6", "…"] }
}

keys is the full list for the requested channel — 41 keys for a call agent, 9 for a whatsapp agent, 46 with no bot_type — trimmed above to three. notes carries the five standing caveats, preceded, for any key whose allowlist runs past 60 values, by a line saying allowed_values was truncated and that POST /validate-config still checks against the whole list.

For an agent that speaks (every type except whatsapp, and the unfiltered schema), the response carries three more top-level fields:

FieldMeaning
voice_tiersThe priced voice tiers an agent can run on via config.voice_tier instead of its own STT/LLM/TTS stack — id, label, blurb, price_credits_per_minute, speech_inputs, voices, voices_searchable, languages, multilingual, billing (flat) and minimum_seconds
stt_languagesFor each speech-recognition model, the language codes it can transcribe, in that model's own code format (hi-IN or hi). Fold to the base subtag before comparing with a TTS language list
multilingual{stt: [...], tts: [...]} — the speech-recognition and speech models that can follow the caller's language when language is "multi"

Per-key fields:

FieldMeaning
keyThe config key name, spelled exactly as the runtime reads it
typestring, integer, number, boolean, array or object
channelsWhich channels read this key: voice, text, or both
summaryOne line on what the key does
defaultThe effective value when the key is absent, or null when there is no single default
if_omittedWhat actually happens on a live call when you leave the key out — read this before deciding a key is optional
allowed_valuesThe closed allowlist, or null when the value is free-form
values_depend_onThe other key that decides this key's allowlist — voice depends on tts_model, for example. When this is set, allowed_values alone is not enough; validate the pair
bounds[min, max] for a numeric key, otherwise null
validated_on_writetrue when a bad value is rejected with 422 at write time. When this is false, a bad value is accepted and fails at call time — that is exactly what POST /validate-config is for

minimal_example is a complete, working body for POST /api/v1/bots. Post it as-is and you get an agent that answers a call and speaks.

403 when a cm_ key lacks bots:read.

POST/api/v1/bots/validate-config

Checks a config object and tells you what is wrong with it. It writes nothing, touches no bot, and never returns an error for a bad config — the findings are the answer. Requires bots:read.

FieldTypeRequiredNotes
configobjectYesThe config you are about to write. At most 200 keys
bot_typestring | nullNoNarrows the check to one channel's keys. Omit it to check against every key. Any value outside the five bot types is 422
curl -X POST https://api.callmissed.com/api/v1/bots/validate-config \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "bot_type": "inbound_call",
    "config": {
      "tts_model": "deepgram-aura-2",
      "voice": "shubh",
      "max_duration_seconds": 600,
      "greeting": "Hi, thanks for calling."
    }
  }'
{
  "ok": false,
  "findings": [
    {
      "key": "max_duration_seconds",
      "severity": "warning",
      "code": "unknown_key",
      "message": "max_duration_seconds is not read from an agent's config — it is the name the call runtime receives. Set max_call_duration_seconds instead, or this setting does nothing.",
      "did_you_mean": ["max_call_duration_seconds"],
      "allowed": null
    },
    {
      "key": "voice",
      "severity": "error",
      "code": "incompatible",
      "message": "'shubh' is not a speaker on deepgram-aura-2 — it belongs to bulbul:v3; set tts_model to it. On a standard voice call the session refuses to start; where a softer path applies, the caller hears a different speaker than you chose.",
      "did_you_mean": ["janus"],
      "allowed": ["agathe", "agustina", "alvaro", "ama", "amalthea", "andromeda"]
    }
  ]
}

ok is true only when no finding has severity error. An empty findings array means the config is clean. A clean config returns:

{ "ok": true, "findings": [] }
SeverityMeaning
errorThe call will not work, or will be audibly wrong — a voice the chosen engine cannot speak, a model id that does not exist, a value outside its bounds
warningAccepted, but probably not what you meant — an unknown key, a typo'd tool name, a missing greeting
CodeRaised when
unknown_keyThe key is not one the runtime reads. It is stored and ignored
unknown_valueThe value is not in that key's allowlist
wrong_typeThe value is the wrong JSON type for the key
out_of_rangeA numeric value is outside the key's bounds. It is clamped at call time
incompatibleThe value is invalid given another key — most often a voice that the chosen tts_model cannot speak
missing_recommendedA key you almost certainly want is absent — no greeting, for instance, which leaves the caller with silence

did_you_mean is a list — the closest valid spellings, at most three, and [] when nothing is close enough. allowed is that key's allowlist when the list is closed, capped at 40 values, and null otherwise.

POST /bots, PUT /bots/{id} and PATCH /bots/{id}/config do not reject a config that fails this check. They report it: their response carries a config_warnings array holding the same findings. Read it after every write — an empty array is the only confirmation that what you wrote will actually run. Read paths (GET /bots, GET /bots/{id}) always return config_warnings: []; ask POST /validate-config to check a stored config.

403 when a cm_ key lacks bots:read.

DELETE/api/v1/bots/{bot_id}

Returns 204 with an empty body. Requires bots:write.

curl -X DELETE https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab \
  -H "Authorization: Bearer cm_your_api_key"

403 without bots:write; 404 when not found.

POST/api/v1/bots/{bot_id}/toggle

Flips is_active. No request body. Requires bots:write.

curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/toggle \
  -H "Authorization: Bearer cm_your_api_key"

Returns the bot with the new is_active. 403 without bots:write; 404 when not found.

While is_active is false the agent takes no phone calls: a caller to a number it answers hears a short "Sorry, this number isn't taking calls right now. Please try again later." and the call ends (unless a call menu, call flow, call queue or agent squad on that number answers it), call queues route past it, campaigns that use it pause, and placing a call with its bot_id returns 409. It also stops replying on WhatsApp, Instagram and Facebook (messages still reach the inbox), declines WhatsApp calls to its numbers at once (Call Desk numbers still ring a person), and on a linked personal WhatsApp number it sends nothing, including reminders, while messages keep syncing.

POST/api/v1/bots/{bot_id}/deploy/verify

Live connectivity check against the bot's channel, using credentials stored in the bot's own config. No request body. Requires bots:write; a JWT caller must also be owner or admin.

Which credentials are read depends on the bot type:

Bot typeReads from configChecked against
whatsapp, whatsapp_voicephone_number_id, access_tokenMeta Graph API
inbound_call, outbound_call, ivraccount_sid, auth_tokenTwilio REST API

The result is also persisted back into config as channel_verified, last_verified_at, and error_message.

curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/deploy/verify \
  -H "Authorization: Bearer cm_your_api_key"
{
  "channel_verified": true,
  "last_verified_at": "2026-08-04T11:05:33.129004+00:00",
  "error_message": null
}

A failed check is still 200, with channel_verified: false and a reason such as Missing phone_number_id or access_token, WhatsApp API returned 401, Twilio API returned 401, or Provider API timeout — try again.

StatusCause
403Key missing bots:write, or Only owners/admins can verify deployments for a JWT caller
404Bot not found

POST/api/v1/bots/{bot_id}/mcp-servers/probe

Connects to each MCP server saved in the bot's config.mcp_servers, lists its tools, and reports whether it answered. Read-only: nothing is stored. Requires bots:read.

FieldTypeRequiredConstraints
idsarray of string | nullNoProbe only the servers with these ids. At most 10. Omit the body (or send null) to probe every saved server
curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/mcp-servers/probe \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"ids": ["crm"]}'
{
  "servers": [
    {
      "id": "crm",
      "host": "mcp.example.com",
      "ok": true,
      "tool_count": 3,
      "tools": ["find_contact", "create_lead", "log_call"],
      "error": null
    }
  ]
}
FieldMeaning
idThe server's id from config.mcp_servers
hostHostname only. The full URL and any header value you stored are never returned
oktrue when the server answered tools/list
tool_count / toolsHow many tools it advertised, and their names (the list is capped at 40)
errorOn failure, the class of error only (for example ConnectError or TimeoutException), otherwise null

An unreachable server is still 200, with ok: false. An id that is not saved on the bot is simply absent from servers.

StatusCause
403Key missing bots:read
404Bot not found
422More than 10 ids

Versions

Stage edits to an agent's system_prompt and config as a draft, publish them atomically, and roll back to any earlier published version. The bot row is always what serves live traffic: a draft changes nothing about what the agent says until you commit it.

  • Version 1 is created lazily. The first time you create a draft on a bot, a committed Baseline version holding the current live prompt and config is created first, so your first commit has something to roll back to. A bot that has never been versioned lists [].
  • One open draft per bot. A second POST returns 409 naming the existing draft.
  • Committed versions are immutable. Rollback never rewrites history: it appends a new committed version carrying the old snapshot and publishes it.
  • History is capped at 50 committed versions per bot. After each commit or rollback the oldest committed versions beyond that are deleted; a draft is never pruned.
  • Writes made outside versioning — PUT /api/v1/bots/{bot_id}, PATCH .../config, PUT .../prompt — still change the live agent directly and do not create a version.

The version object:

FieldTypeMeaning
iduuidVersion row id
tenant_id, bot_iduuidOwner and agent
version_numberinteger1, 2, 3… per bot. Every path below addresses a version by this number
labelstring | nullYour name for the version
system_promptstringThe prompt snapshot
configobject | nullThe config snapshot, credential-redacted exactly like the bot's own config
statusstringdraft or committed
created_by, committed_byuuid | nullDashboard user who created / committed it; null when done with an API key
commit_messagestring | nullSet on commit; Rollback to vN on a rollback version
created_at, updated_at, committed_atdatetimecommitted_at is null on a draft

GET/api/v1/bots/{bot_id}/versions

Every version of the bot, newest first. Requires bots:read.

curl https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/versions \
  -H "Authorization: Bearer cm_your_api_key"
[
  {
    "id": "0f6c2a51-3d1e-4b7a-9c55-1f2e3d4c5b6a",
    "tenant_id": "a0b1c2d3-4455-6677-8899-aabbccddeeff",
    "bot_id": "b1f2c3d4-5678-90ab-cdef-1234567890ab",
    "version_number": 2,
    "label": "Shorter greeting",
    "system_prompt": "You are a concise support agent for Acme.",
    "config": { "greeting": "Hi, Acme here." },
    "status": "draft",
    "created_by": null,
    "committed_by": null,
    "commit_message": null,
    "created_at": "2026-09-20T10:00:00Z",
    "updated_at": "2026-09-20T10:02:00Z",
    "committed_at": null
  },
  {
    "id": "8b1d7e40-6a2c-4f13-a8e9-2c3d4e5f6a7b",
    "tenant_id": "a0b1c2d3-4455-6677-8899-aabbccddeeff",
    "bot_id": "b1f2c3d4-5678-90ab-cdef-1234567890ab",
    "version_number": 1,
    "label": "Baseline",
    "system_prompt": "You are a friendly support agent for Acme.",
    "config": { "greeting": "Hi, thanks for calling Acme. How can I help?" },
    "status": "committed",
    "created_by": null,
    "committed_by": null,
    "commit_message": "Baseline snapshot of the live agent",
    "created_at": "2026-09-20T10:00:00Z",
    "updated_at": "2026-09-20T10:00:00Z",
    "committed_at": "2026-09-20T10:00:00Z"
  }
]

404 Bot not found.

GET/api/v1/bots/{bot_id}/versions/{version_number}

One version. Requires bots:read. 404 Bot not found or Version not found.

POST/api/v1/bots/{bot_id}/versions

Stages a draft. Every field defaults to the bot's current live value, so an empty body {} snapshots what is live and lets you edit it. Returns 201 with the draft. Requires bots:write.

FieldTypeRequiredConstraints
labelstring | nullNo1-255 chars
system_promptstring | nullNoMax 20000 chars. Defaults to the live prompt
configobject | nullNoThe whole config for this version (not a merge). Defaults to the live config. The same sub-keys as POST /api/v1/bots are validated
curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/versions \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"label": "Shorter greeting", "system_prompt": "You are a concise support agent for Acme."}'

A config you send replaces the live config wholesale when committed, and the live config's credential fields (for example access_token) are not readable back through the API. Omit config to inherit the stored config, credentials included, rather than rebuilding it from a redacted read.

StatusCause
404Bot not found
409A draft already exists (vN) — edit or commit it first
422A field outside its bounds, or a malformed validated config sub-key

PATCH/api/v1/bots/{bot_id}/versions/{version_number}

Edits the open draft. Omitted fields are unchanged; config, when sent, replaces the draft's config. Same field table as create. Requires bots:write.

StatusCause
404Bot not found or Version not found
409Committed versions are immutable — roll back to it instead

POST/api/v1/bots/{bot_id}/versions/{version_number}/commit

Publishes the draft: copies its system_prompt and config onto the live agent and marks the version committed, in one transaction — a call in progress reads either the whole old agent or the whole new one. Requires bots:write.

FieldTypeRequiredConstraints
commit_messagestring | nullNoMax 2000 chars
curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/versions/2/commit \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"commit_message": "Tighter greeting after call review"}'

Returns the committed version.

StatusCause
404Bot not found or Version not found
409Version is already committed

POST/api/v1/bots/{bot_id}/versions/{version_number}/rollback

Rolls back by moving forward: appends a new committed version carrying version version_number's snapshot, with commit_message Rollback to vN, and publishes it. The target version is untouched. No request body. Returns 201 with the new version. Requires bots:write.

curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/versions/1/rollback \
  -H "Authorization: Bearer cm_your_api_key"
StatusCause
404Bot not found or Version not found
409Only a committed version can be rolled back to — commit the draft instead

Knowledge base

Entries attached to one bot. See Knowledge Base for retrieval behaviour.

GET/api/v1/bots/{bot_id}/knowledge

Lists entries for the bot, newest first. Requires knowledge:read.

curl https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/knowledge \
  -H "Authorization: Bearer cm_your_api_key"
[
  {
    "id": "k9e8d7c6-b5a4-4321-9876-543210fedcba",
    "bot_id": "b1f2c3d4-5678-90ab-cdef-1234567890ab",
    "name": "Refund policy",
    "content": "Refunds are issued within 7 business days...",
    "metadata": { "source": "handbook" },
    "file_url": null,
    "file_size_bytes": null,
    "format": null,
    "status": "indexed",
    "error_message": null,
    "created_at": "2026-07-11T09:00:00Z"
  }
]

403 without knowledge:read; 404 Bot not found.

POST/api/v1/bots/{bot_id}/knowledge

Adds a text entry. Returns 201. Requires knowledge:write.

FieldTypeRequiredConstraints
namestringYes1-255 chars
contentstringYes1-100000 chars
metadataobject | nullNoFree-form JSON
curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/knowledge \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Refund policy",
    "content": "Refunds are issued within 7 business days of the return being received.",
    "metadata": { "source": "handbook" }
  }'

Returns the created entry. 403 without knowledge:write; 404 Bot not found; 422 when content is empty or over 100000 chars.

POST/api/v1/bots/{bot_id}/knowledge/upload

Uploads a document as multipart/form-data and extracts its text synchronously. Returns 201. Requires knowledge:write.

PartTypeRequiredConstraints
filefileYesExtension must be pdf, docx, or txt. Max 20 MB (20,971,520 bytes)
curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/knowledge/upload \
  -H "Authorization: Bearer cm_your_api_key" \
  -F "[email protected]"
{
  "id": "k1a2b3c4-d5e6-4f70-8192-a3b4c5d6e7f8",
  "bot_id": "b1f2c3d4-5678-90ab-cdef-1234567890ab",
  "name": "handbook.pdf",
  "content": "Acme employee handbook\n\n1. Refunds...",
  "metadata": null,
  "file_url": null,
  "file_size_bytes": 184203,
  "format": "pdf",
  "status": "indexed",
  "error_message": null,
  "created_at": "2026-08-04T11:12:00Z"
}

Check status on the response. When extraction yields nothing usable the entry is still created with status: "failed" and an error_message such as No text content could be extracted. The name is the uploaded filename.

StatusCause
400File exceeds 20MB limit, or Unsupported format: <ext>. Use PDF, DOCX, or TXT.
403Key missing knowledge:write
404Bot not found

DELETE/api/v1/bots/{bot_id}/knowledge/{entry_id}

Returns 204 with an empty body. Requires knowledge:write.

curl -X DELETE https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/knowledge/k9e8d7c6-b5a4-4321-9876-543210fedcba \
  -H "Authorization: Bearer cm_your_api_key"

403 without knowledge:write; 404 Knowledge entry not found when the entry is not on that bot in your tenant.

Agent documents

Files an agent can send to a customer on WhatsApp when asked, such as a trip itinerary, a brochure or a price list. Turn on the send_document tool on the agent (see Voice agent tools); the agent passes the document's name and the file arrives in the chat. Free-form files deliver only inside WhatsApp's 24-hour customer-service window, so outside it the agent sends an approved template first.

GET/api/v1/bots/{bot_id}/documents

Lists the agent's documents, oldest first. Requires bots:read.

curl https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/documents \
  -H "Authorization: Bearer cm_your_api_key"
[
  {
    "id": "d3a4b5c6-7e8f-4a90-b1c2-d3e4f5a6b7c8",
    "name": "itinerary",
    "filename": "Goa-5-day-itinerary.pdf",
    "content_type": "application/pdf",
    "byte_size": 238114,
    "created_at": "2026-10-09T10:15:00Z"
  }
]

403 without bots:read; 404 Bot not found.

POST/api/v1/bots/{bot_id}/documents

Uploads a file as multipart/form-data. Returns 201. Requires bots:write.

PartTypeRequiredConstraints
filefileYesPDF, JPEG or PNG, checked by the file's contents rather than its extension or Content-Type. Max 20 MB (20,971,520 bytes)
namestringNoUp to 64 chars. The name the agent picks the file by, stored lower-case with punctuation turned into spaces (Trip Itinerary! becomes trip itinerary). Defaults to the filename without its extension

An agent holds up to 20 documents, and each name is unique per agent.

curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/documents \
  -H "Authorization: Bearer cm_your_api_key" \
  -F "[email protected]" \
  -F "name=itinerary"
{
  "id": "d3a4b5c6-7e8f-4a90-b1c2-d3e4f5a6b7c8",
  "name": "itinerary",
  "filename": "Goa-5-day-itinerary.pdf",
  "content_type": "application/pdf",
  "byte_size": 238114,
  "created_at": "2026-10-09T10:15:00Z"
}

filename is what the customer sees on WhatsApp.

StatusCause
400File exceeds 20MB limit; Unsupported file. Use PDF, JPEG or PNG.; Give the document a name, e.g. itinerary; This agent already has a document named '<name>'; or An agent can hold up to 20 documents
403Key missing bots:write
404Bot not found
503The file could not be stored, try again
503File storage is not configured

DELETE/api/v1/bots/{bot_id}/documents/{document_id}

Removes the document and its file. Returns 204 with an empty body. Requires bots:write. Deleting the agent removes its documents too.

curl -X DELETE https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/documents/d3a4b5c6-7e8f-4a90-b1c2-d3e4f5a6b7c8 \
  -H "Authorization: Bearer cm_your_api_key"

403 without bots:write; 404 Document not found when the document is not on that agent in your tenant.

Callback calls

Calls a WhatsApp customer asked the agent for, arranged by its request_callback tool (WhatsApp callback calls).

GET/api/v1/bots/{bot_id}/callbacks

Lists the agent's callbacks, newest first. Requires bots:read.

QueryTypeDefaultConstraints
limitinteger501-200
curl "https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/callbacks?limit=20" \
  -H "Authorization: Bearer cm_your_api_key"
[
  {
    "id": "e4f5a6b7-c8d9-4e0f-a1b2-c3d4e5f6a7b8",
    "to_e164": "+919812345678",
    "status": "pending",
    "due_at": "2026-10-09T11:30:00Z",
    "attempts": 0,
    "reason": "Wants to discuss the Goa package",
    "error": null,
    "telephony_call_id": null,
    "call_status": null,
    "created_at": "2026-10-09T08:02:11Z"
  }
]
FieldNotes
statuspending (waiting for due_at), dialing, placed, failed or cancelled
attemptsTimes a call was tried
errorWhy it failed, was cancelled, or is being retried
telephony_call_id, call_statusThe placed call and its status (initiated, ringing, in_progress, completed, failed, no_answer, busy)

403 without bots:read; 404 Bot not found.

POST/api/v1/bots/{bot_id}/callbacks/{callback_id}/cancel

Cancels a callback that has not started dialing and returns it. Requires bots:write. Cancelling one that is already cancelled returns it unchanged.

curl -X POST https://api.callmissed.com/api/v1/bots/b1f2c3d4-5678-90ab-cdef-1234567890ab/callbacks/e4f5a6b7-c8d9-4e0f-a1b2-c3d4e5f6a7b8/cancel \
  -H "Authorization: Bearer cm_your_api_key"
StatusCause
403Key missing bots:write
404Callback not found
409Only a callback that hasn't started dialing can be cancelled

On this page