Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.usecustory.com/llms.txt

Use this file to discover all available pages before exploring further.

Actions are what the automation does after the trigger fires. Custory supports a small set of high-leverage actions. That constraint is useful. It keeps automations focused on workflows that actually move customer-context work forward instead of becoming unreadable pipelines.

​
Action limits and validation

Custory automations currently support:
  • Up to 3 actions per automation
  • At most 1 Update journey action in a single automation
That is a good boundary for smaller teams. If a workflow needs more than three major steps, it usually needs simplification first.

​
Action types

Custory currently supports these action types:
  • Update journey
  • Get GitHub PRs
  • Get analytics
  • Send Slack message
  • Send Discord message
  • Create GitHub issue
  • Create Jira issue
  • Create Linear issue

​
Update journey

This action asks Custory AI to update the scoped journey directly from the trigger context. Use it when the workflow should keep the journey current without manual rewriting. Best for:
  • GitHub merge follow-up
  • Analytics-driven journey refresh
  • Structured maintenance after important product changes
Requirements:
  • Clear AI instructions
  • The automation scoped to exactly one journey
Good instructions tell AI what to update, what to avoid duplicating, and when to skip changes.

​
Get analytics

This action fetches a structured analytics snapshot from:
  • PostHog
  • Stripe
Use it when the automation needs fresh product or revenue signal before taking the next step.

​
PostHog presets

Supported presets:
  • Event volume change
  • Top events
Event volume change requires:
  • Event name
  • Window between 1 and 90 days
Top events supports:
  • Window between 1 and 90 days
  • Top-events limit

​
Stripe presets

Supported presets:
  • Subscription activity
  • Payment failures
  • Refund activity
This action is especially powerful when followed by:
  • Update journey
  • Send Slack or Discord message
  • Create a task
The analytics step gives downstream actions structured context to work from.

​
Get GitHub PRs

This action fetches pull request context from selected GitHub repositories. Use it when the automation should inspect code activity before notifying the team or updating the journey. Configuration includes:
  • GitHub integration
  • Repositories
  • PR state: open, closed, or all
  • Since mode
  • Optional base branches
  • Limit
If using Last N days, the PR window must be between 1 and 30 days. This action is useful as a context step before:
  • Journey updates
  • Team summaries
  • Validation prompts

​
Send Slack message

Sends a message to a Slack channel. Use it when the automation should make the signal visible to the team where they already operate. Configuration includes:
  • Slack integration
  • Channel
  • Message mode
Message modes:
  • AI drafted
  • Static template
Use AI-drafted messages when the content depends on context and should be readable. Use static templates when the format is highly predictable.

​
Send Discord message

Sends a message to a Discord channel. Configuration is similar to Slack, but Discord uses a channel ID rather than a channel name-style entry. Use it when the team or community-heavy workflow already lives in Discord.

​
Create GitHub issue

Creates a GitHub issue from the trigger context. Configuration includes:
  • GitHub integration
  • Owner
  • Repo
  • Target mode
  • Title mode
  • Body mode
  • Optional labels
  • Optional AI instructions

​
Target mode

Task-creation actions support:
  • Selected step
  • Trigger step
Use Selected step when the automation should always write back to a specific journey location. Use Trigger step when the workflow should follow the step of the incoming event automatically.

​
Text modes

Issue title and body can be:
  • AI drafted
  • Static template
AI mode is generally better when the issue should include customer context, expected outcome, and a more thoughtful handoff.

​
Create Jira issue

Creates a Jira issue from the trigger context. Configuration includes:
  • Jira integration
  • Project
  • Issue type
  • Optional priority
  • Target mode
  • Summary mode
  • Description mode
  • Optional AI instructions
Use this when delivery lives inside Jira and the team needs structured handoff from customer context into project execution.

​
Create Linear issue

Creates a Linear issue from the trigger context. Configuration includes:
  • Linear integration
  • Team
  • Optional numeric priority
  • Target mode
  • Title mode
  • Description mode
  • Optional AI instructions
This is often the fastest route from opportunity to execution for lean product teams.

​
AI drafted vs static templates

Several actions support both AI-drafted and static output. Choose AI drafted when:
  • The wording should adapt to context
  • The handoff quality matters
  • The automation is transforming messy signal into readable output
Choose static template when:
  • The message structure is fixed
  • The team wants exact phrasing
  • The action is mostly mechanical

​
How to combine actions well

The strongest patterns usually follow one of these shapes:

​
Fetch -> update -> notify

Example:
  • Get analytics
  • Update journey
  • Send Slack message

​
Trigger -> create task -> notify

Example:
  • Item priority threshold crossed
  • Create Jira issue
  • Send Slack message

​
Delivery signal -> update -> validate

Example:
  • GitHub PR merged
  • Update journey
  • Send Discord message

​
Common mistakes

​
Using too many creation actions in one workflow

If a single trigger opens several external tasks at once, the team often loses track of which artifact actually matters.

​
Writing weak AI instructions

Make a summary is too vague. Better instructions specify audience, desired outcome, and what context to include.

​
Using selected step when trigger step is the real intent

If the automation should follow the step where the event happened, hard-coding a selected step can quietly send work to the wrong place.