Skip to main content

Agents

A Strongly agent is a persistent AI assistant with its own memory, rules, and tools. You chat with it through the Strongly UI, your own apps, or the API. The platform handles deployment, lifecycle, and scale -- you focus on what the agent should do.

Two ways to build an agent​

PathBest for
Agent Builder wizardNew agents authored from scratch. Pick a model, write a personality, wire in tools, seed starter content.
Promote from workflowYou already built a workflow with an agent node. Promote it to a deployable agent.

Both paths produce the same kind of agent. The wizard is faster for new work; promotion is the right call when you have an existing workflow that already does what you want.

What you configure​

SettingWhat it controls
ModelWhich AI model powers the agent, plus optional fallbacks the runtime promotes if the primary is unhealthy.
CapabilitiesWhat the agent is allowed to do.
PersonalityThe agent's voice, tone, and style.
Operating PromptThe long-form instructions for how the agent should use its tools, memory, and rules.
Context PolicyHow the agent fits long conversations into the model's context window.
HeartbeatThe agent's autonomous background loop: when enabled, it wakes on a cron schedule, in the timezone you pick (your own by default), and works its due heartbeat tasks.
Session PolicyIdle timeout, max session length, concurrent-session limits per user.
Connect data sources and toolsThe data sources, add-ons, and tools the agent can reach, and the workflows and add-ons attached to it (see below).
ChannelsWhere the agent is reachable (for example a website embed).
KnowledgeThe knowledge bases the agent answers from (created in Library > Knowledge), and shared pools the agent also reads. Click Apply changes so a running agent picks a change up.

Each setting has its own section in Agent Settings once the agent is deployed -- open it from the gear icon on the agent's row. Memory, rules, skills, tasks, and preferences seeded in the wizard are editable at any time from the Library.

Agent Settings

Attached workflows and add-ons​

Besides its connectors, an agent can have workflows it calls as tools (attached when a workflow is promoted to an agent) and add-ons provisioned for it by the Agent Builder. They are listed under Attached workflows and add-ons in Agent Settings > Connect data sources and tools, each with its type and status.

While an agent lists a workflow or add-on, it is in use: deleting it is refused and the message names the agent. Click Remove next to it to take it off the agent. Nothing is deleted: the workflow or add-on stays where it is, and you can delete it afterwards from its own page. If the agent is running, click Apply changes so it restarts without it.

Reference templates​

Three reference templates, built and maintained by Strongly, can pre-fill the wizard: Iris (personal assistant), Atlas (research assistant), and Sage (support assistant). Walk the wizard, review the pre-filled values, and edit anything before submit. Ready-made packaged agents are also available to install from the Marketplace.

Operating an agent​

After deploy, each agent has its own page in the Strongly UI with:

  • Chat -- talk to the agent.
  • Sessions -- see who's chatting right now, end / pause / transfer sessions.
  • Tasks, Threads, Logs -- agent activity and history.
  • Settings -- change any setting from the build process. A running agent reads most settings when it starts: after you save one, Apply changes in the header stays highlighted, even if you leave the page and come back, until you click it to restart the agent on the new settings. It is off while a tab has unsaved edits: save the tab first, since a restart uses only saved settings. The session policy, and edits to the text of the agent's operating prompt, apply without it.

After a platform update a running agent is redeployed on the new version for you. If that is refused for the user who started it, Settings shows a restart warning until it is redeployed. See Platform Updates.

Next steps​