Post-Call Actions

OverviewCopy Anchor Link

After a call ends, call analysis writes a summary, the caller's sentiment, the outcome and the answers your agent's analysis_variables ask for (for example star_rating, callback_requested). Post-call actions act on those answers automatically.

Each agent has an ordered list of rules. A rule is a set of conditions (all must pass) and one or more actions. A rule with no conditions runs after every analysed call.

Actions run in the background shortly after the analysis finishes. Each action runs at most once per call, its status is recorded per call, and temporary failures are retried with backoff.

AuthenticationCopy Anchor Link

Authorization: Bearer cm_your_api_key

Reading rules, options, recipes, test results and run history needs the bots:read scope. Creating, changing or deleting rules, and the on/off switches, need bots:write.

All paths below are under https://api.callmissed.com/api/v1/bots/{bot_id}/post-call.

RecipesCopy Anchor Link

curl https://api.callmissed.com/api/v1/bots/{bot_id}/post-call/recipes \
  -H "Authorization: Bearer cm_your_api_key"

Ready-made rules, already filled in from this agent's own analysis variables. Each recipe has a key, a plain-language title and description, available (with unavailable_reason when not), needs (the blanks you still choose) and the rule it would save.

Save one with its blanks filled:

curl -X POST https://api.callmissed.com/api/v1/bots/{bot_id}/post-call/rules/from-recipe \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{ "recipe": "low_rating_followup",
        "fills": { "assignee_user_id": "6f1c…", "notify_user_id": "6f1c…" } }'

Building blocksCopy Anchor Link

curl https://api.callmissed.com/api/v1/bots/{bot_id}/post-call/options \
  -H "Authorization: Bearer cm_your_api_key"

Returns everything a rule can use for this agent: variables (its analysis variables with their types), call_fields, operators per type, tokens for text, team, contact_fields, sendable whatsapp_templates with their inputs, and shared Google sheets. Use these ids; a rule that names anything else is refused.

RulesCopy Anchor Link

curl -X POST https://api.callmissed.com/api/v1/bots/{bot_id}/post-call/rules \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Low rating follow-up",
    "conditions": [ { "field": "star_rating", "op": "lte", "value": 2 } ],
    "actions": [
      { "type": "create_task",
        "config": { "title": "Low rating ({{star_rating}}) from {{caller_number}}",
                    "description": "{{summary}}", "due_in_hours": 24 } },
      { "type": "notify",
        "config": { "subject": "Low rating from {{caller_number}}", "message": "{{feedback_comment}}" } }
    ]
  }'

A rule is checked when it is saved: every condition must name one of the agent's analysis variables (or a call field) with an operator that suits its type, every {{token}} must exist, and every team member, contact field, template and spreadsheet must belong to your workspace. A problem returns 422 with a message saying what to fix.

ConditionsCopy Anchor Link

{ "field": "...", "op": "...", "value": ... }. field is one of the agent's analysis variable names, or call.sentiment, call.disposition, call.goal_met or call.direction.

An unanswered variable never matches a comparison, only is_empty. If a variable is later removed from the agent, conditions on it stop matching.

ActionsCopy Anchor Link

{ "type": "...", "config": { ... } }. Text fields take {{tokens}}: any analysis variable, plus summary, sentiment, disposition, goal_met, caller_number, contact_name, agent_name, call_date, call_time, call_duration, call_direction and call_link.

WhatsApp consent. A message is never sent to a number that opted out of your WhatsApp messages or is on your do-not-contact list. A marketing template is sent only to contacts who agreed to receive WhatsApp messages. The action is recorded as skipped, with the reason, when either applies. WhatsApp messages are billed as usual.

Test a ruleCopy Anchor Link

curl -X POST https://api.callmissed.com/api/v1/bots/{bot_id}/post-call/dry-run \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{ "session_id": "…" }'

Runs the agent's saved rules (or one rule_id, or an unsaved rule) against a past call of this agent. For each rule: matched, each condition with the call's actual value and passed, and for each action a plain-language would and a blocked_reason when it would be held back. Nothing is sent, saved or created. GET /recent-calls?limit=20 (1-50) lists the agent's latest analysed calls to pick from. A call without a finished analysis returns 409.

Run historyCopy Anchor Link

curl "https://api.callmissed.com/api/v1/bots/{bot_id}/post-call/runs?session_id=…" \
  -H "Authorization: Bearer cm_your_api_key"

One entry per action per call, newest first (limit 1-100, default 50; offset 0-10000): rule_name, action_type, status (pending, running, succeeded, failed, skipped), attempts, error (why it failed or was skipped) and result (for example task_id, ticket_id, contact_id, send_id).

From an AI assistantCopy Anchor Link

The Account MCP server can set all of this up: list_post_call_recipes, get_post_call_options, create_post_call_rule_from_recipe, create_post_call_rule, update_post_call_rule, toggle_post_call_rule, set_post_call_actions_enabled, delete_post_call_rule, list_post_call_recent_calls, dry_run_post_call_rule and list_post_call_action_runs. The post_call_actions playbook (get_playbook) walks an assistant through interviewing you first.