Hyphen
Log In

Agent Capabilities

Explore the comprehensive capabilities of Hyphen Agent, a powerful tool that combines conversational AI with task-based operations for enhanced automation and insights.

Hyphen Agent combines conversational answers with task-based operations work. Availability can depend on your organization permissions, connected integrations, and the resources Agent can access.

Table of Contents

Task Guides

For deeper task-specific guidance, see:

Deployments

Kubernetes

Activity Answers

Agent can answer questions about organization activity from recorded Hyphen events. It supports listing, counting, finding, and summarizing activity across projects, apps, environments, deployments, preview deployments, teams, members, feature flags, segments, links, domains, ENV, and Agent tasks.

Example questions:

  • "What happened in this project today?"
  • "How many deployments failed this week?"
  • "Who created this environment?"
  • "What has Jordan changed recently?"
  • "Summarize the last 24 hours of production events."

If a resource name matches multiple resources, Agent asks for clarification instead of guessing.

Chat Sessions and Slack

Agent chat sessions support ongoing organization conversations in the Hyphen App. When Slack is connected, you can also DM the Hyphen bot or mention it in a channel thread.

Slack chat supports thread continuity, Hyphen resource links, Slack mentions, and input requests for confirmations or follow-up questions.

Member Memory

Agent can remember durable member preferences when you explicitly ask it to remember, update, forget, or clear something. Memory is useful for preferences such as answer format, common environment names, or recurring context.

Member memory is not a secret store. Do not save credentials, tokens, private keys, or sensitive personal data in memory.

Log Analysis

Agent can analyze deployment logs and surface findings by severity. By default, Hyphen Agent analyzes available cloud provider container logs. If the app is connected to New Relic APM, Agent can use APM data instead.

Log analysis results can include:

  • Log coverage metrics.
  • Findings grouped by category and severity.
  • GitHub issue actions, such as created, commented, or skipped.
  • Repository and pull request context when a fix is proposed.

Deployment Optimizer

Agent can investigate project environment deployment and runtime signals, run optimizer simulations, and recommend changes such as right-sizing, consolidation, or shutting down underused resources.

Optimizer recommendations are intended to preserve declared availability expectations. When an optimizer run requires approval, Hyphen presents an input request before Agent proceeds.

Stale Feature Flag Cleanup

Agent can identify Toggle flags with no recent usage or unchanged return values over a 90-day window. When the relevant app has a connected GitHub repository, Agent can open a cleanup pull request to remove stale flag references from code.

Perpetual feature flags are excluded from stale flag cleanup.

GitHub Pull Request Automation

When GitHub is connected, Agent can connect findings to GitHub issues and pull requests. For log analysis, Agent can create a new issue, comment on a matching existing issue, or open a pull request with a proposed fix.

Agent-created code changes are reviewable. Agent does not merge pull requests automatically as part of this workflow.

Kubernetes Optimizer

Agent can analyze a workload in a registered Kubernetes cluster and recommend findings or replica guidance from inventory evidence such as current replicas, ready replicas, related pods, services, warning events, sibling workloads, planned events, and autoscaling context.

Kubernetes optimizer recommendations are review-first. Scale recommendations require approval before anything changes, and Agent skips manual replica guidance when matching HorizontalPodAutoscaler evidence is detected.

Kubernetes Resource Cleanup

Agent can scan registered Kubernetes clusters for stale, dead, or unused resources and present cleanup candidates for review. Kubernetes resource cleanup runs inspect cluster inventory, apply the configured stale threshold and ignore lists, and create approval requests before deleting any resource.

Kubernetes cleanup policies can be managed at the organization level, and individual clusters can override the organization policy when a cluster needs its own schedule, threshold, or ignore list.

Deployment Controls

Agent can inspect project environment deployment settings and answer deployment readiness questions from chat. You can ask what is configured for an environment, which apps and cloud targets are included, whether the deployment is ready to run, or what the latest non-preview deployment run looked like.

Agent can also start a configured deployment or preview deployment. Before creating a deployment run, Agent checks readiness and presents a confirmation request with the target project, environment, preview context, and selected builds. If builds are not specified, Agent uses the latest app build for non-preview deployments and the latest preview build for preview deployments.

Follow-up deployment requests preserve the prior deployment target, including preview deployment context, so requests like "deploy again" can reuse the last confirmed project and environment.

Deployment Resource Cleanup

Agent can detect unused or stale deployment resources created or tracked by Hyphen, including orphaned containers and expired preview deployment resources. Deployment resource cleanup work is presented through Agent runs so you can review status, output, and any required approvals.

From chat, Agent can help with project, short link, and QR code operations.

Supported project actions include listing projects and creating a project after collecting required details and confirmation.

Supported link and QR code actions include reading short links, creating short links, deleting short links, and helping create, retrieve, or delete QR codes for existing short links.

Scheduled and Recurring Work

Agent tasks can be scheduled or recurring. Task and run detail views show timing, status, related resources, conversation history, and execution output.

Hyphen also surfaces deployment settings drift on project environment views when current deployment settings differ from the latest run snapshot. Agent optimizer work can use deployment and runtime context, but the deployment drift alert itself is documented as a Hyphen Deploy surface rather than an Agent task.