Skip to content
Changelog

Report a bug

Managed Agent Automation Examples

Each Type.com automation example is a complete automation that you can copy into src/main.ts and deploy. Every page has the same parts: what it does, how it runs, the full code, how it works, a test, and the values to replace.

The code is written to be read top to bottom. The stages are written inline, in the order that they run. Each stage starts with a comment that says what it does.

Each example has a graph. Each node in the graph is one key in the code, and its shape shows its type:

entry next onResult hourly schedule check code review agent record code
Trigger Code stage Agent stage
ShapeNodeIn the code
PillTriggerA key in deployment.subscriptions
RectangleCode stageA code({ ... }) key in stages
HexagonAgent stageAn agent({ ... }) key in stages

The arrows use the words from the code: entry is the subscription’s entry stage, next is a next(...) decision, and onResult is where an agent’s result goes.

Type.com runs these automations on its own production systems and engineering channels. The examples are simplified versions, with placeholder IDs.

ExampleTriggersWhat it doesShows you how to
Monitor production securityEvery 15 minutesChecks the database, request logs, and Sentry for attacks. An agent investigates new signals and escalates in Slack.Read several sources, dedupe with a checkpoint, and call an agent only when needed
Turn bug reports into fixesSlack message, Slack reaction, scheduleFiles a Linear story for each bug post, and an agent opens a fix pull request. A ✅ closes the story, and Mondays remind owners.Run three triggers in one automation
Verify, merge, and watch bug fixesSlack message, scheduleA verifier on a different model checks the fix video before the pull request opens. Low-risk fixes merge when green, and the deploy is watched with backoff.Chain two agents on different models, merge safely, and monitor after a deploy
Fix one security finding a dayDaily scheduleFinishes an open fix, or audits the code and fixes the most severe finding. Reports in Slack.Route a run between two agents that share one contract
Investigate errors on requestSlack reactionA 🔍 on a bug report starts an investigation with the team’s skill. Code posts the root cause.Start an agent on demand and use a skill
ExampleTriggerStagesShows you how to
Diagnose CI failuresGitHub workflow runcode → agent → codeSkip noise with an AI question, then investigate with an agent
Write daily release notesCron schedulecode → agent → codeUse code, the sandbox, an agent, and a skill together
Triage Slack requests with an agentSlack messagecode → agent → codeGive one step to the managed agent and use its typed result
Triage inbound emailEmailcode → agent → codeBranch on an agent’s decision
Poll an API for new itemsInterval schedulecode → codeRead a Custom API and remember progress with a checkpoint
Receive webhooks from any serviceWebhook URLcodeAccept HTTP events from any service

An instruction automation needs no code. Ask Type in a Space, and say when it runs, what it reads, what it makes, and where the result goes:

JobAsk Type
Morning support digest“Every weekday at 8:00, read the support tickets opened since yesterday, group them by theme with the /support-triage skill, and post a summary in this Space.”
Weekly blocker sweep“Every Friday at 15:00, list the Linear issues in the current cycle that are blocked or have not changed in five days, and post them with their owners.”
Pull request review“When a pull request is opened in acme/project, review it with the /pr-review skill and comment on the pull request.”
Linear triage“When a Linear issue is created, add a label, suggest an owner, and ask for missing reproduction steps.”
Competitor watch“When the competitor’s changelog feed has a new entry, summarize what changed and why it matters to us.”
Vendor inbox“For each email to this Space’s address, summarize it and post it only if it needs a reply.”

Type.com proposes the trigger, the schedule, the destination, and the connections before it makes the automation. See the Quickstart.

The examples use these placeholder values. Replace them with your own:

PlaceholderWhat it isHow to find yours
ch_exampleA Space IDtype-cli spaces list --json
orgint_example, orgint_slack, orgint_linear, …A connection IDtype-cli integrations list --json
ches_exampleAn event source ID (webhook, email, or RSS)source.sourceId in the automation that Type made for the source
C0123456789A Slack channel IDThe channel details in Slack
acme/projectA GitHub repositoryowner/repo