Automations
Run AI prompts on a schedule or after specific events
Automations run the Modem Agent without a chat. Give one a prompt and say when it runs: on a schedule, or when an event happens in a connected tool. An automation can also skip the agent and send each event to a webhook.
How it works
An automation is a prompt plus a run condition. In the dashboard you pick a Type:
| Type | Use it for | Configuration |
|---|---|---|
| Scheduled | Digests, recurring checks, one-time follow-ups | Pick Daily, Repeats, or Once |
| Trigger | Reacting to PR or MR activity, topic changes, Stripe billing, or new feedback | Pick one or more events (up to 20) and, if you want, a condition |
An automation has the same tools as the agent in chat. Depending on which integrations you’ve connected and which tools you leave on, a run can post to Slack, send email, create Linear issues, link topics to GitHub PRs, or take other actions for you.
For a trigger automation, Modem hands the event’s details to the agent with your prompt. You don’t need template variables or placeholders. Write what you want done.
Outcomes
The Then row says what a run does:
| Then | What happens |
|---|---|
| Run the Modem agent | The default. The agent runs your prompt with the event’s details and the tools you leave on. |
| Send a webhook | Modem POSTs the event to an HTTPS URL you choose, signed with Standard Webhooks, and runs no agent. See Webhooks. |
Automation types
Scheduled automations
Use a scheduled automation for a recurring or one-time run.
Run context
Each scheduled run starts with a short run context: when the automation last ran successfully, what changed since then, whether your organization has been used lately, which integrations are connected, and where output goes by default. It is delivered as the schedule.tick event:
- Run the Modem agent: the agent reads it before your prompt. A prompt such as “Summarize what’s new since the last run” works without dates, because the agent knows when the last run was.
- Send a webhook: the delivery carries it as
event, withnameset toschedule.tickand the run context asdata. See Webhooks.
Run Once gets the same run context as a run at the scheduled time.
The counts cover a window from the automation’s previous successful run up to this one, and a Run Once counts as a run. A first run looks back 30 days, and so does a run whose previous success is older than that. Skipped and failed runs don’t move the window, so the next run still covers everything since the last successful one.
The run context holds counts and yes/no answers, never message text, names, or people. The agent, or your webhook’s receiver, looks up the details itself.
| Field | Meaning |
|---|---|
automation.id |
The automation’s id, the same on every run |
automation.run_id |
This run’s id |
schedule.kind |
cron, rrule, every, or at |
schedule.slot_at |
When this run started (UTC) |
schedule.timezone |
The IANA timezone the schedule is written in |
schedule.previous_run_at |
When the automation’s previous successful run started, or null if it has never completed one |
window.from, window.to |
The span the counts below cover; to is this run |
ingest.activity |
some if Modem received at least one event from your sources in the window, else none |
ingest.last_event_at |
When the newest of those events arrived, or null |
topics.activity |
some if topics.created or topics.updated is above 0, else none |
topics.created |
Active topics created in the window. Like the Topics page, it leaves out topics that are no longer active or have nothing left in them, so a topic counted once can drop out later |
topics.updated |
Active topics created before the window that gained new evidence in it |
organization.activity |
some if, in the 30 days before this run, Modem received an event, someone talked to the agent outside automations, or this automation was created; else none |
integrations.connected |
The slugs of your connected integrations, such as slack or linear, sorted |
outputs.default |
Your organization’s default output destination as { "channel": "slack" } (email, slack, discord, or msteams), or null if none is set |
A full example is under Scheduled run.
If Modem can’t build the run context, an agent run goes ahead without it. A webhook run fails instead and sends nothing, so your receiver never gets a scheduled delivery without its event.
Skip a scheduled run with nothing to do
A scheduled run can check its run context before the agent starts or the webhook is sent, and skip itself. A skipped run costs nothing and appears in the run history as skipped, with the reason. These checks apply to runs at the scheduled time; Run Once always runs.
- A filter is an expression over
ingest.activity,topics.activity, andorganization.activity, using the same syntax as a trigger filter."ingest.activity == 'some'"skips a digest when nothing arrived since the last one. - A condition is a plain-language sentence, like a trigger condition: “at least three topics were created since the last run”.
When you save a condition, Modem decides once how to check it:
| The condition needs | How it is checked | A run that doesn’t match costs |
|---|---|---|
| Only the run context’s fields | A fast evaluation model checks it before the agent starts, and skips the run | Nothing |
| Anything else, such as what customers said (“a customer reported a billing bug”) | Every run starts, and the agent checks the condition first | One run |
To skip runs for free, write the condition in terms of the run context: activity, topic counts, connected integrations, or time since the last run.
Set either one through the API or by asking the agent; the dashboard doesn’t edit them yet, and keeps them when you change a schedule’s time without changing its type. In the API, add filter or condition to a cron, rrule, every, or at schedule:
{
"type": "cron",
"expr": "0 9 * * 1",
"filter": "ingest.activity == 'some' || topics.activity == 'some'"
}
A filter works with either outcome. A condition needs the agent outcome, so an automation that sends a webhook can’t have one.
Trigger automations
A trigger automation runs as soon as one of its events fires.
| Source | Events |
|---|---|
| Modem | Topic Created, Topic Priority High, Group Added to Topic, Group Updated on Topic, Feedback Created |
| Conversations | Chat Activity, Call Activity, Support Ticket Activity, Feedback Activity, Social Post Activity |
| GitHub | GitHub PR Created, GitHub PR Merged, GitHub PR Dismissed |
| GitLab | GitLab MR Created, GitLab MR Merged, GitLab MR Closed |
| Stripe | Stripe Customer Created, Stripe Subscription Created, Stripe Subscription Canceled, Stripe Trial Ending, Stripe Payment Failed |
| Sentry | Sentry Feedback Created |
| GitHub, GitLab | Pull Request Activity |
| Issue trackers | Issue Activity |
| Modem | Topic Activity |
The dashboard offers the activity events, Topic Created, Topic Priority High, and the Stripe events. The other events are superseded by an activity event plus a filter: Pull Request Activity with change.kind == 'merged' replaces GitHub PR Merged, for example. An automation already on one of the old events keeps it. Stripe events appear in the picker only once Stripe is connected.
Topic Created and Topic Priority High fire for every topic in your organization, whatever started it, and can’t be limited to channels or products. To react only to conversations in certain channels, use a conversation activity event with a placement == 'created' or topic.priority_change == 'became_high' filter instead.
Group Added to Topic and Group Updated on Topic currently fire only for X conversations, and an automation on them must be limited to specific channels or products.
Feedback Created fires for every submission to an in-app feedback channel, for user feedback captured through Sentry, and for posts captured through Canny. It carries the message, the reporter’s contact details when given, the channel, any tags, and a provider field naming the source. Limit the automation to one feedback channel to react to a single app or agent.
Conversation activity events
The five conversation activity events fire each time Modem sorts a conversation into a topic, or decides it belongs to none. There is one per tab of the Sources page, and the kind of the conversation’s source container decides which one fires. A conversation from a custom source fires the event for its container’s kind: chat_channel fires Chat Activity, support_thread and email_thread fire Support Ticket Activity, and feedback_submission fires Feedback Activity.
| Event | Fires for | Adds a block with |
|---|---|---|
Chat Activity (chat.activity) |
Slack, Discord, Microsoft Teams | chat: the workspace, whether the conversation is a thread, and whether the channel is shared with another organization |
Call Activity (call.activity) |
Gong, Fathom | call: the title, start time, length, recording link, summary, and speakers |
Support Ticket Activity (support_ticket.activity) |
Intercom, Zendesk, Front, Plain, Pylon, Jira Service Management, email | support_ticket: the subject, status, priority, assignee, requester, tags, the channel it arrived through, dates, and satisfaction rating |
Feedback Activity (feedback.activity) |
In-app feedback channels, Sentry, Canny | feedback: the title, page, tags, release, environment, trace and replay IDs, and the board, status, and votes |
Social Post Activity (social_post.activity) |
X | social_post: whether it is a post, reply, quote, or repost, the search it matched, its engagement, and the author’s followers |
Every conversation activity event also carries:
- The conversation’s recent messages, each with its author’s name, handle, email, and role: staff, customer, external, or bot. For a call, the messages are the turns Modem used as evidence rather than the end of the transcript, so each is marked
cited. - The people and customer company involved.
- The topic and the products.
An automation on one of these events must be limited to specific channels or products, or carry a filter or trigger condition. A channel limit can only name channels from the event’s built-in sources, so reach a custom source with a product limit, a filter, or a condition.
Pull request and issue activity
Pull Request Activity fires when a GitHub pull request or GitLab merge request is opened, marked ready for review, merged, closed without merging, or reopened. New commits, label changes, and other updates don’t fire it. The event carries the pull request with its description, labels, and newest comments; who made the change and each person’s role (staff, customer, external, or bot); the customer company involved; and the topic and products it concerns.
Issue Activity fires when a GitHub, GitLab, Linear, or Jira issue is opened, closed, reopened, moved to another status (Linear and Jira), or gets a new comment. Edits, label changes, and comment edits don’t fire it. The event carries the issue with its status, priority, and newest comments, plus the same people, company, topic, and product details as a pull request.
An automation on either event must be limited to specific repositories or products, or carry a filter or trigger condition. Linear and Jira issues reach only automations that use a filter or trigger condition and no repository or product limit.
Topic activity
Topic Activity (topic.activity) fires when Modem creates a topic (change.kind is created) or moves an existing topic to another priority level (priority_changed, with the old level in change.from_priority_level). A topic created in an update reports only created, and a topic built only from messages older than 30 minutes (a first import of history) reports nothing. A rescore that keeps the level, or that runs on evidence older than 30 minutes, fires nothing, and neither does changing a topic’s priority by hand.
The event carries the topic with its priority level and the scorer’s reasoning, how many customer companies, customer people, other outside people, and staff wrote on it, the main customer company, and its products. It carries no message text.
A topic belongs to no channel, so an automation on Topic Activity can’t be limited to channels or products and must carry a filter or trigger condition. Topic Created and Topic Priority High still fire as before.
Filters
A filter narrows an activity trigger to the events you care about. Modem matches it in code, before any trigger condition or agent run, so an event the filter rules out costs nothing.
An activity trigger can’t share an automation with other kinds of events. You can select several of the five conversation activity events together; Pull Request Activity, Issue Activity, and Topic Activity each run alone.
In the dashboard, pick a preset under When, such as “a customer sends a message” or “a pull request is merged”, then narrow it with + Filter. It offers only the fields every watched event has. The conversation presets watch all five conversation activity events; to watch one, pick its own preset, such as “a support ticket has new activity”.
Through the API, set filter to one expression of field == 'value' and field in ['a', 'b'] clauses joined with && and ||, with parentheses:
{
"type": "event",
"events": ["support_ticket.activity"],
"filter": "source.name in ['zendesk', 'intercom'] && author.role == 'customer'"
}
| Event | Filter fields |
|---|---|
| Chat Activity, Call Activity, Feedback Activity | source.name, placement, author.role, topic.priority_change, topic.priority_level, topic.from_priority_level, audience |
| Support Ticket Activity | source.name, placement, author.role, topic.priority_change, topic.priority_level, topic.from_priority_level, audience, support_ticket.status |
| Social Post Activity | source.name, placement, author.role, topic.priority_change, topic.priority_level, topic.from_priority_level, audience, social_post.kind |
| Pull Request Activity | source.name, change.kind, base.ref, audience |
| Issue Activity | source.name, change.kind, audience |
| Topic Activity | change.kind, topic.priority_level, change.from_priority_level |
| Field | Values |
|---|---|
source.name |
On a conversation activity event, one of that event’s sources from the table in Trigger automations, or a custom source’s name. On Pull Request Activity, github or gitlab. On Issue Activity, github, gitlab, linear, or jira. |
placement |
created, joined, moved, unchanged, or none. |
author.role |
The role of whoever wrote the newest message: staff, customer, external, bot, or unknown. |
topic.priority_change |
became_high when this update pushed the topic to high priority, else none. |
topic.priority_level |
The topic’s priority after this update: very_low, low, default, high, or very_high. |
topic.from_priority_level |
The topic’s priority before this update, when this update changed it. null when this update did not change the priority or its evidence is older than 30 minutes, so it matches no value. A topic never scored before counts as default. |
support_ticket.status |
open, in_progress, waiting, done, or canceled. A ticket with no status matches no clause on it. |
social_post.kind |
post, reply, quote, or repost. |
change.kind |
On Pull Request Activity, created, ready_for_review, merged, closed, or reopened. On Issue Activity, created, closed, reopened, status_changed, or commented. On Topic Activity, created or priority_changed. |
change.from_priority_level |
On Topic Activity, the topic’s priority before a priority_changed: very_low, low, default, high, or very_high. null for created, so it matches no value. |
base.ref |
The branch a pull request merges into. Match a name exactly with == or in, such as base.ref in ['main', 'release/v2'], or use an RE2 regex with base.ref.matches('^release/'). A regex matches anywhere in the name unless you anchor it with ^ and $. |
audience |
internal or public, by who can see where the item lives. The rules are below. |
To run when a topic becomes very high priority, compare the two priority fields:
topic.priority_level == 'very_high' && topic.from_priority_level in ['very_low', 'low', 'default', 'high']
On Topic Activity, include topics created at that level. To run when a topic is created very high or rises to very high:
topic.priority_level == 'very_high' && (change.kind == 'created' || change.from_priority_level in ['very_low', 'low', 'default', 'high'])
To run when a topic is created high or very high, or rises into either:
topic.priority_level in ['high', 'very_high'] && (change.kind == 'created' || change.from_priority_level in ['very_low', 'low', 'default'])
A filter with || has no dashboard preset; the When row shows it as written.
How Modem decides audience:
| Where the item lives | Audience |
|---|---|
| A Slack channel of your own workspace, or a Slack direct message | internal |
| A Slack Connect channel | public |
A Discord channel @everyone can’t view |
internal |
A Discord channel @everyone can view |
public |
| A GitHub pull request or issue | Its repository’s visibility; private and GitHub Enterprise internal repositories count as internal |
| A Linear issue | Always internal. Modem doesn’t reflect a Linear issue’s public share link. |
| A support ticket, social post, feedback, call, Microsoft Teams, GitLab, Jira, or a channel Modem hasn’t synced | public. Anything Modem can’t confirm is internal is public. |
The agent uses the same rule when wording a reply, except in Microsoft Teams, where DMs and synced channels not shared outside your organization count as internal.
Target branches
Target branches applies only to automations still on the GitHub PR created, merged, and dismissed events, which the picker no longer offers. For Pull Request Activity, use a base.ref filter instead.
It takes names or patterns separated by spaces, such as main release/*, matched case-sensitively against the branch the PR targets. * matches any characters, including /, so release/* matches release/team/v2; no other wildcard syntax is supported. Leave it empty to match every branch.
Through the API, include targetBranches in the event schedule:
{
"type": "event",
"events": ["github.pr.merged"],
"targetBranches": ["main"]
}
Trigger conditions
A trigger condition says in plain language when the automation should run, in up to 2000 characters. Modem checks each event against it before each run. Leave it out to run on every event.
On an activity trigger, choose + Smart trigger under the When preset; it opens an and row for the condition. Other triggers have an Only run when… field.
Write one full sentence that names its subject and can be answered from the event itself:
- the issue is a bug report with no assignee
- the conversation reports that a feature is broken or failing
- the customer asks about pricing, billing, or a plan change
- the pull request changes the public API
On an activity trigger, the line under the condition says as you type whether it can be answered from the event alone or needs more.
Hover the event pill in the prompt editor to see an example of the details a condition is judged against. The full examples for the activity events are under Example payloads.
When you save the condition, Modem decides once how to check it:
| The condition needs | How it is checked | A non-matching event costs |
|---|---|---|
| Only the event’s details | A fast evaluation model checks each event before any agent runs | Nothing |
| History, counts over time, or data outside the event (“this customer has reported this twice this month”) | The automation runs on every matching event, and the agent checks the condition first | One run |
Either way the agent sees the condition in its prompt and does nothing when the event doesn’t satisfy it.
A condition also counts as a scope. An event that must be limited to channels or products, such as Support Ticket Activity, can carry a condition instead. You can combine them; the channel or product filter applies first.
The automation’s detail page lists Trigger evaluations: each recent event with what Modem decided, and counts for the last seven days.
| Decision | Meaning |
|---|---|
| Ran | The condition held and the automation ran. The row shows how likely the condition was judged to hold; 50% or more runs the automation. |
| Skipped | The condition did not hold, or could not be decided from the event. |
| Agent decides | The automation ran, and the agent checked the condition first. |
| Unavailable | Modem could not check the condition, so the automation did not run. |
| Too large | The event’s details were too large to check, so the automation did not run. |
Through the API, include condition in the event schedule. Modem sets conditionMode on the saved schedule to record how it checks the condition (judge or agent); leave it out of what you send.
{
"type": "event",
"events": ["support_ticket.activity"],
"condition": "a customer reports that a feature is broken or failing"
}
Example payloads
These are the examples the event pill shows for each activity event, exactly as Modem publishes them, followed by a scheduled run’s context. A condition can name any field in them in plain language. A webhook delivers the same object as event.data.
Chat Activity
{
"placement": "joined",
"source": {
"name": "slack",
"container_kind": "chat_channel"
},
"conversation": {
"key": "slack:1760541902.118200",
"url": "https://northwind.slack.com/archives/C07NWSHARED/p1760542260118900",
"channel": {
"id": "dddddddd-1111-4000-8000-000000000001",
"name": "ext-northwind"
},
"message_count": 2,
"messages": [
{
"posted_at": "2026-10-15T15:25:02.000Z",
"url": "https://northwind.slack.com/archives/C07NWSHARED/p1760541902118200",
"author": {
"name": "Priya Raman",
"handle": "priya.raman",
"email": "priya@northwind.example",
"role": "customer",
"person_id": "p-northwind-priya",
"company_name": "Northwind"
},
"text": "Our order webhooks have been timing out since about 2pm. Nothing changed on our side.",
"cited": true
},
{
"posted_at": "2026-10-15T15:31:00.000Z",
"url": "https://northwind.slack.com/archives/C07NWSHARED/p1760542260118900",
"author": {
"name": "Tomás Ortega",
"handle": "tomas",
"email": "tomas@modem.example",
"role": "staff",
"person_id": "p-staff-tomas",
"company_name": null
},
"text": "Thanks Priya, looking now. Can you share one delivery id that timed out?",
"cited": true
}
]
},
"author": {
"name": "Tomás Ortega",
"handle": "tomas",
"email": "tomas@modem.example",
"role": "staff",
"person_id": "p-staff-tomas",
"company_name": null
},
"audience": "public",
"group": {
"id": "sg-0101",
"summary": "Northwind reports order webhooks timing out since 2pm with no change on their side.",
"last_activity_at": "2026-10-15T15:31:00.000Z",
"message_count": 2
},
"topic": {
"id": "t-webhook-timeouts",
"title": "Webhook deliveries time out",
"summary": "Several customers report webhook deliveries timing out this afternoon.",
"issue_type": "bug_report",
"priority": 50,
"lifecycle_state": "open",
"is_new": false,
"group_count": 3,
"priority_change": "none",
"priority_level": "high",
"from_priority_level": "very_high"
},
"people": [
{
"person_id": "p-northwind-priya",
"name": "Priya Raman",
"handle": "priya.raman",
"email": "priya@northwind.example",
"role": "customer",
"company": {
"id": "c-northwind",
"name": "Northwind",
"domain": "northwind.example"
}
},
{
"person_id": "p-staff-tomas",
"name": "Tomás Ortega",
"handle": "tomas",
"email": "tomas@modem.example",
"role": "staff",
"company": null
}
],
"customer": {
"id": "c-northwind",
"name": "Northwind",
"domain": "northwind.example"
},
"products": [
{
"id": "prod-webhooks",
"name": "Webhooks"
}
],
"chat": {
"workspace": "Modem",
"is_thread": true,
"is_shared": true
}
}
Call Activity
{
"placement": "created",
"source": {
"name": "gong",
"container_kind": "transcript"
},
"conversation": {
"key": "gong:7782214409123",
"url": null,
"channel": null,
"message_count": 214,
"messages": [
{
"posted_at": "2026-10-14T17:12:40.000Z",
"url": null,
"author": {
"name": "Dana Whitfield",
"handle": null,
"email": "dana@lumenlabs.example",
"role": "customer",
"person_id": "p-lumen-dana",
"company_name": "Lumen Labs"
},
"text": "The main thing for us is the rate limit. Our nightly sync hits it every run and we end up retrying for an hour.",
"cited": true
},
{
"posted_at": "2026-10-14T17:13:21.000Z",
"url": null,
"author": {
"name": "Dana Whitfield",
"handle": null,
"email": "dana@lumenlabs.example",
"role": "customer",
"person_id": "p-lumen-dana",
"company_name": "Lumen Labs"
},
"text": "Per-record. We would move to bulk if it supported filters.",
"cited": true
}
]
},
"author": {
"name": "Dana Whitfield",
"handle": null,
"email": "dana@lumenlabs.example",
"role": "customer",
"person_id": "p-lumen-dana",
"company_name": "Lumen Labs"
},
"audience": "public",
"group": {
"id": "sg-0202",
"summary": "Lumen Labs hits the API rate limit on every nightly sync and would move to bulk export if it supported filters.",
"last_activity_at": "2026-10-14T17:13:21.000Z",
"message_count": 2
},
"topic": {
"id": "t-rate-limit-sync",
"title": "Nightly syncs hit the API rate limit",
"summary": "Customers syncing record by record run into the API rate limit and ask for filtered bulk export.",
"issue_type": "feature_request",
"priority": 0,
"lifecycle_state": "open",
"is_new": true,
"group_count": 1,
"priority_change": "none",
"priority_level": "default",
"from_priority_level": "low"
},
"people": [
{
"person_id": "p-lumen-dana",
"name": "Dana Whitfield",
"handle": null,
"email": "dana@lumenlabs.example",
"role": "customer",
"company": {
"id": "c-lumen",
"name": "Lumen Labs",
"domain": "lumenlabs.example"
}
}
],
"customer": {
"id": "c-lumen",
"name": "Lumen Labs",
"domain": "lumenlabs.example"
},
"products": [
{
"id": "prod-api",
"name": "Public API"
}
],
"call": {
"title": "Lumen Labs renewal check-in",
"started_at": "2026-10-14T16:58:00.000Z",
"duration_seconds": 2710,
"recording_url": "https://app.gong.io/call?id=7782214409123",
"summary": "Renewal check-in. Lumen Labs is happy with reporting but blocked by API rate limits on their nightly sync.",
"participants": [
{
"name": "Sam Okafor",
"email": "sam@modem.example",
"affiliation": "internal"
},
{
"name": "Dana Whitfield",
"email": "dana@lumenlabs.example",
"affiliation": "external"
},
{
"name": "Ravi Menon",
"email": "ravi@lumenlabs.example",
"affiliation": "external"
}
]
}
}
Support Ticket Activity
{
"placement": "unchanged",
"source": {
"name": "intercom",
"container_kind": "support_thread"
},
"conversation": {
"key": "intercom:conv:88213",
"url": "https://app.intercom.com/a/inbox/acme/conversation/88213",
"channel": {
"id": "dddddddd-2222-4000-8000-000000000001",
"name": "Intercom inbox"
},
"message_count": 3,
"messages": [
{
"posted_at": "2026-10-15T14:02:11.000Z",
"url": null,
"author": {
"name": "Megan Kowalski",
"handle": null,
"email": "megan@acme.example",
"role": "customer",
"person_id": "p-acme-megan",
"company_name": "Acme"
},
"text": "Our CSV export has been failing since this morning. Every time I click Export on the Contacts page it spins and then shows 'Export failed (500)'. We need this for a board report today.",
"cited": true
},
{
"posted_at": "2026-10-15T14:09:40.000Z",
"url": null,
"author": {
"name": "Greg Haddad",
"handle": null,
"email": "greg@modem.example",
"role": "staff",
"person_id": "p-staff-greg",
"company_name": null
},
"text": "Sorry about that, Megan. Roughly how many contacts are in the list you are exporting, and does a smaller filtered export work?",
"cited": false
},
{
"posted_at": "2026-10-15T14:15:03.000Z",
"url": null,
"author": {
"name": "Megan Kowalski",
"handle": null,
"email": "megan@acme.example",
"role": "customer",
"person_id": "p-acme-megan",
"company_name": "Acme"
},
"text": "About 48,000. A filter down to 2,000 exports fine, so it looks like a size problem.",
"cited": true
}
]
},
"author": {
"name": "Megan Kowalski",
"handle": null,
"email": "megan@acme.example",
"role": "customer",
"person_id": "p-acme-megan",
"company_name": "Acme"
},
"audience": "public",
"group": {
"id": "sg-0001",
"summary": "Acme reports the Contacts CSV export failing with a 500 for a 48k-row list; a 2k-row filtered export works.",
"last_activity_at": "2026-10-15T14:15:03.000Z",
"message_count": 2
},
"topic": {
"id": "t-export-500",
"title": "Large contact exports fail with 500",
"summary": "Customers exporting large contact lists hit a 500; small exports succeed.",
"issue_type": "bug_report",
"priority": 50,
"lifecycle_state": "open",
"is_new": false,
"group_count": 2,
"priority_change": "became_high",
"priority_level": "high",
"from_priority_level": "default"
},
"people": [
{
"person_id": "p-acme-megan",
"name": "Megan Kowalski",
"handle": null,
"email": "megan@acme.example",
"role": "customer",
"company": {
"id": "c-acme",
"name": "Acme",
"domain": "acme.example"
}
},
{
"person_id": "p-staff-greg",
"name": "Greg Haddad",
"handle": null,
"email": "greg@modem.example",
"role": "staff",
"company": null
}
],
"customer": {
"id": "c-acme",
"name": "Acme",
"domain": "acme.example"
},
"products": [
{
"id": "prod-contacts",
"name": "Contacts"
}
],
"support_ticket": {
"id": "88213",
"subject": "CSV export failing",
"status": "waiting",
"provider_status": "open",
"priority": null,
"assignee": null,
"requester": {
"name": "Megan Kowalski",
"handle": null,
"email": "megan@acme.example",
"role": "customer",
"person_id": "p-acme-megan",
"company_name": "Acme"
},
"tags": ["export", "billing-plan:enterprise"],
"via": null,
"created_at": "2026-10-15T14:02:11.000Z",
"updated_at": "2026-10-15T14:15:03.000Z",
"satisfaction": null
}
}
Feedback Activity
{
"placement": "joined",
"source": {
"name": "feedback-submission",
"container_kind": "feedback_submission"
},
"conversation": {
"key": "feedback-submission:fs_01JAB7Q4ZK",
"url": null,
"channel": {
"id": "dddddddd-3333-4000-8000-000000000001",
"name": "In-app feedback"
},
"message_count": 1,
"messages": [
{
"posted_at": "2026-10-16T09:41:27.000Z",
"url": null,
"author": {
"name": "Ines Duarte",
"handle": null,
"email": "ines@brightline.example",
"role": "customer",
"person_id": "p-brightline-ines",
"company_name": "Brightline"
},
"text": "The revenue chart on the dashboard just spins forever. Other charts load fine.",
"cited": true
}
]
},
"author": {
"name": "Ines Duarte",
"handle": null,
"email": "ines@brightline.example",
"role": "customer",
"person_id": "p-brightline-ines",
"company_name": "Brightline"
},
"audience": "public",
"group": {
"id": "sg-0303",
"summary": "Brightline reports the dashboard revenue chart never finishes loading while other charts work.",
"last_activity_at": "2026-10-16T09:41:27.000Z",
"message_count": 1
},
"topic": {
"id": "t-revenue-chart-spinner",
"title": "Revenue chart never loads",
"summary": "The dashboard revenue chart stays on its loading spinner for some accounts.",
"issue_type": "bug_report",
"priority": 50,
"lifecycle_state": "open",
"is_new": false,
"group_count": 4,
"priority_change": "none",
"priority_level": "high",
"from_priority_level": "very_high"
},
"people": [
{
"person_id": "p-brightline-ines",
"name": "Ines Duarte",
"handle": null,
"email": "ines@brightline.example",
"role": "customer",
"company": {
"id": "c-brightline",
"name": "Brightline",
"domain": "brightline.example"
}
}
],
"customer": {
"id": "c-brightline",
"name": "Brightline",
"domain": "brightline.example"
},
"products": [
{
"id": "prod-dashboards",
"name": "Dashboards"
}
],
"feedback": {
"title": "Revenue chart stuck loading",
"page_url": "https://app.example.com/dashboards/revenue",
"surface": "widget",
"tags": [
{
"key": "plan",
"value": "growth"
}
],
"release": "web@2026.10.15",
"environment": "production",
"trace_id": "4f1c2b9e8a7d4c3b9f0e1d2c3b4a5968",
"replay_id": null,
"board": null,
"status": null,
"votes": null
}
}
Social Post Activity
{
"placement": "created",
"source": {
"name": "x",
"container_kind": "social_feed"
},
"conversation": {
"key": "x:1846021173300441088",
"url": "https://x.com/kaiwrites/status/1846021173300441088",
"channel": {
"id": "dddddddd-4444-4000-8000-000000000001",
"name": "Mentions of our CLI"
},
"message_count": 1,
"messages": [
{
"posted_at": "2026-10-15T19:04:51.000Z",
"url": "https://x.com/kaiwrites/status/1846021173300441088",
"author": {
"name": "Kai Morgan",
"handle": "kaiwrites",
"email": null,
"role": "external",
"person_id": null,
"company_name": null
},
"text": "Excited to try this, but the install script fails on Apple Silicon with \"unsupported architecture\". Anyone else?",
"cited": true
}
]
},
"author": {
"name": "Kai Morgan",
"handle": "kaiwrites",
"email": null,
"role": "external",
"person_id": null,
"company_name": null
},
"audience": "public",
"group": {
"id": "sg-0404",
"summary": "A developer reports the CLI install script failing on Apple Silicon with \"unsupported architecture\".",
"last_activity_at": "2026-10-15T19:04:51.000Z",
"message_count": 1
},
"topic": {
"id": "t-cli-apple-silicon",
"title": "CLI install fails on Apple Silicon",
"summary": "The CLI install script rejects Apple Silicon machines as an unsupported architecture.",
"issue_type": "bug_report",
"priority": 0,
"lifecycle_state": "open",
"is_new": true,
"group_count": 1,
"priority_change": "none",
"priority_level": "default",
"from_priority_level": "very_low"
},
"people": [
{
"person_id": null,
"name": "Kai Morgan",
"handle": "kaiwrites",
"email": null,
"role": "external",
"company": null
}
],
"customer": null,
"products": [
{
"id": "prod-cli",
"name": "CLI"
}
],
"social_post": {
"kind": "quote",
"query": "\"modem cli\" OR @modemdev -is:retweet",
"metrics": {
"likes": 14,
"reposts": 2,
"replies": 5,
"quotes": 1,
"views": 3120,
"measured_at": "2026-10-15T19:06:00.000Z"
},
"author_followers": 8420
}
}
Pull Request Activity
{
"change": {
"kind": "merged",
"occurred_at": "2026-10-16T09:42:10.000Z",
"actor": {
"name": "Sam Okafor",
"handle": "sokafor",
"email": null,
"role": "staff"
}
},
"source": {
"name": "github",
"container_kind": "pull_request"
},
"pull_request": {
"key": "github:2318890451",
"url": "https://github.com/acme-dev/app/pull/4821",
"number": 4821,
"title": "Stream large contact exports instead of buffering them",
"body": "Exports over ~20k rows were built in memory and hit the worker limit, returning a 500. This streams rows to the CSV writer.\n\nFixes the export failures Acme reported.",
"state": "merged",
"is_draft": false,
"audience": "internal",
"repository": {
"full_name": "acme-dev/app"
},
"base_ref": "main",
"head_ref": "sam/stream-exports",
"labels": ["bug", "exports"],
"author": {
"name": "Greg Haddad",
"handle": "ghaddad",
"email": null,
"role": "staff"
},
"assignees": ["ghaddad"],
"reviewers": ["sokafor"],
"created_at": "2026-10-15T16:05:00.000Z",
"merged_at": "2026-10-16T09:42:10.000Z",
"closed_at": "2026-10-16T09:42:10.000Z",
"recent_comments": [
{
"posted_at": "2026-10-16T09:30:44.000Z",
"url": "https://github.com/acme-dev/app/pull/4821#pullrequestreview-3001",
"author": {
"name": "Sam Okafor",
"handle": "sokafor",
"email": null,
"role": "staff"
},
"text": "Verified a 60k-row export locally. Looks good."
}
]
},
"topic": {
"id": "t-export-500",
"title": "Large contact exports fail with 500",
"summary": "Customers exporting large contact lists hit a 500; small exports succeed.",
"issue_type": "bug_report",
"priority": 72,
"lifecycle_state": "in_progress",
"is_new": false,
"group_count": 3
},
"people": [
{
"person_id": "p-staff-greg",
"name": "Greg Haddad",
"handle": "ghaddad",
"email": null,
"role": "staff",
"company": null
},
{
"person_id": "p-staff-sam",
"name": "Sam Okafor",
"handle": "sokafor",
"email": null,
"role": "staff",
"company": null
}
],
"customer": null,
"products": [
{
"id": "prod-contacts",
"name": "Contacts"
}
]
}
Issue Activity
{
"change": {
"kind": "commented",
"occurred_at": "2026-10-16T09:42:10.000Z",
"actor": {
"name": "Tom Becker",
"handle": "tbecker",
"email": "tom@acme.example",
"role": "customer"
},
"from_status": null,
"to_status": null,
"comment": {
"posted_at": "2026-10-16T09:42:10.000Z",
"author": {
"name": "Tom Becker",
"handle": "tbecker",
"email": "tom@acme.example",
"role": "customer"
},
"text": "Still failing for us on the 40k-row export this morning. Is there a workaround?"
}
},
"source": {
"name": "linear",
"container_kind": "issue"
},
"issue": {
"key": "linear:6f1c2d3e-4a5b-4c6d-8e7f-90a1b2c3d4e5",
"url": "https://linear.app/acme/issue/ENG-42/large-contact-exports-fail-with-500",
"identifier": "ENG-42",
"title": "Large contact exports fail with 500",
"body": "Exports over ~20k rows return a 500. Reported by Acme and two other customers.",
"state": "open",
"status": "In Progress",
"status_category": "started",
"priority": "High",
"labels": ["bug", "exports"],
"author": {
"name": "Greg Haddad",
"handle": "greg",
"email": null,
"role": "staff"
},
"assignees": ["Sam Okafor"],
"audience": "internal",
"repository": {
"name": "Data Platform"
},
"created_at": "2026-10-14T16:05:00.000Z",
"closed_at": null,
"comment_count": 3,
"recent_comments": [
{
"posted_at": "2026-10-15T11:20:00.000Z",
"url": "https://linear.app/acme/issue/ENG-42#comment-1a2b",
"author": {
"name": "Sam Okafor",
"handle": "sam",
"email": null,
"role": "staff"
},
"text": "Reproduced with a 25k-row list. Working on streaming the CSV writer."
},
{
"posted_at": "2026-10-16T09:42:10.000Z",
"url": "https://linear.app/acme/issue/ENG-42#comment-3c4d",
"author": {
"name": "Tom Becker",
"handle": "tbecker",
"email": "tom@acme.example",
"role": "customer"
},
"text": "Still failing for us on the 40k-row export this morning. Is there a workaround?"
}
]
},
"topic": {
"id": "t-export-500",
"title": "Large contact exports fail with 500",
"summary": "Customers exporting large contact lists hit a 500; small exports succeed.",
"issue_type": "bug_report",
"priority": 72,
"lifecycle_state": "in_progress",
"is_new": false,
"group_count": 3
},
"people": [
{
"person_id": "p-tom",
"name": "Tom Becker",
"handle": "tbecker",
"email": "tom@acme.example",
"role": "customer",
"company": {
"id": "c-acme",
"name": "Acme",
"domain": "acme.example"
}
},
{
"person_id": "p-staff-greg",
"name": "Greg Haddad",
"handle": "greg",
"email": null,
"role": "staff",
"company": null
},
{
"person_id": "p-staff-sam",
"name": "Sam Okafor",
"handle": "sam",
"email": null,
"role": "staff",
"company": null
}
],
"customer": {
"id": "c-acme",
"name": "Acme",
"domain": "acme.example"
},
"products": [
{
"id": "prod-contacts",
"name": "Contacts"
}
]
}
Topic Activity
{
"change": {
"kind": "priority_changed",
"occurred_at": "2026-11-12T16:20:00.000Z",
"from_priority_level": "high"
},
"topic": {
"id": "t-duplicate-orders",
"title": "Duplicate orders created at checkout",
"summary": "A slow checkout submit creates two orders for one payment; customers refund the duplicates by hand.",
"issue_type": "bug_report",
"priority": 100,
"priority_level": "very_high",
"priority_reasoning": "Two customer accounts report duplicate orders from a single payment this week, and both are refunding by hand. Money is charged twice, so the impact is direct.",
"lifecycle_state": "open",
"is_new": false,
"group_count": 3
},
"customer": {
"id": "c-halcyon",
"name": "Halcyon Freight",
"domain": "halcyonfreight.com"
},
"reach": {
"customer_companies": 2,
"customer_people": 3,
"external_people": 0,
"staff_people": 1
},
"products": [
{
"id": "prod-checkout",
"name": "Checkout"
}
]
}
Scheduled run
The run context of a weekly digest in a busy organization, a week after its last run.
{
"automation": {
"id": "0193a1f2-6c4e-7b1a-9d3e-5f2a8c7b4e10",
"run_id": "7c9e6679-7425-50de-944b-e07fc1f90ae7"
},
"schedule": {
"kind": "cron",
"slot_at": "2026-11-09T09:00:00.000Z",
"timezone": "Europe/Berlin",
"previous_run_at": "2026-11-02T09:00:00.000Z"
},
"window": {
"from": "2026-11-02T09:00:00.000Z",
"to": "2026-11-09T09:00:00.000Z"
},
"ingest": {
"activity": "some",
"last_event_at": "2026-11-09T08:57:12.000Z"
},
"topics": {
"activity": "some",
"created": 4,
"updated": 11
},
"organization": {
"activity": "some"
},
"integrations": {
"connected": ["github", "linear", "slack", "zendesk"]
},
"outputs": {
"default": {
"channel": "slack"
}
}
}
Creating an automation
The quickest way is to ask the agent:
The agent writes the prompt and picks the automation type for you.
Creating an automation manually
Navigate to Automations
Go to Automations in your dashboard.
Create a new automation
Click Create Automation and choose Create manually. Give it a name and write the prompt you want the agent to run.
Choose the automation type
Pick Trigger or Scheduled.
Configure when it runs
If you chose Scheduled, pick:
- Daily for runs on selected days at a set time
- Repeats for runs every 30 minutes to 24 hours, on the days you select
- Once for a single future run
If you chose Trigger, pick a preset or event under When and, if you want, narrow it with a filter or a trigger condition. To wait before running, set a Delay of up to 7 days under More Options.
Choose what runs
Under Then, keep Run the Modem agent, or choose Send a webhook and enter the URL to deliver each event to. A webhook automation has no prompt, tools, or outputs.
Choose tools
Under More Options, the Tools row lists every connected tool, all on by default. Turn off any you don’t want the automation to use.
Activate
Save the automation. A scheduled automation runs at its configured time. A trigger automation runs when one of its events fires.
Managing automations
The automations list shows each automation’s type, enabled status, recent runs, and last updated time. Scheduled automations show when they run; trigger automations show the events they listen for.
From the list you can:
- Pause and resume an automation without deleting it
- Run Once to start a scheduled automation now. Trigger automations only run when their event fires.
- Edit the prompt, type, or tools
- View run history to see past runs, their results, and any errors
- View trigger evaluations on an automation with a condition, to see which events ran it and which it skipped