Skip to content

Managed agents#

Autonomous agents created directly on Lens Agents run Prism Agent, configured through conversation. They learn, remember, and operate on their own — within the boundaries you set. Lens Agents provisions the agent through the built-in Prism Agent sandbox template; the rest of what a managed agent does is Prism Agent runtime behavior, documented under Prism Agent.

  • Creating a managed agent

    Provision a managed agent from the built-in Prism Agent sandbox template.

  • Heartbeat monitoring

    Natural-language monitoring instructions with smart alert suppression.

  • Scheduling

    Recurring tasks via cron, intervals, or one-shot timers.

  • Memory

    Two-layer memory system: searchable database plus always-in-context summary.

  • Agent identity

    The markdown documents that define each agent's identity, behavior, and context.

  • Subagents

    Delegate larger tasks to a Claude Code sub-agent running in the workspace.

  • Web chat

    Interact with your agent through the Prism Agent web UI.

  • Slack

    DMs, channels, and multi-user threads — agents as first-class Slack participants.

  • Webhooks

    External systems push events to your agent — Alertmanager, PagerDuty, GitHub, or any JSON webhook.

Design your first managed agent with us

Scoping a managed agent is the most hands-on part of an evaluation. We work through the role (SRE, support, compliance, or specific to your operation), the autonomy level, the policies, the heartbeat cadence, and which channel it lives in. Talk to us with the use case.