Create and Manage Agents

Use Agents to create specialized support agents for different products, docs areas, customer segments, or API domains.

When To Create More Than One Agent

Use one default agent when every customer should receive the same product and API support. Create additional agents when the answer should depend on a different context set, site, product line, or audience.

Pattern	Example
Docs agent	Routes docs visitors to public API documentation and SDK examples.
Product agent	Routes signed-in app users to product-specific help and account-safe flows.
Segment agent	Routes enterprise customers to higher-touch runbooks and escalation rules.
API area agent	Routes billing, webhook, or data-platform questions to focused sources.

Create an Agent

Open Agents

Go to Agents in the workspace sidebar.

Create a new agent

Add a name and description that operators can recognize.

Attach context

Choose one or more sources from Context.

Tune behavior

Review persona, instructions, greeting, confidence threshold, and guardrails.

Save and test

Ask representative support questions before routing real customers to the agent.

Test Before Routing Customers

Use questions that match real support traffic:

Test	Expected outcome
Greeting	Fast, normal response without unnecessary source retrieval.
Capability question	Clear explanation of what the agent can help with.
Known endpoint	Cited answer with the right method, path, fields, and auth requirement.
Request body or error debugging	Grounded answer from source context, or a focused question for missing details.
Missing context	Clarification or handoff instead of invented API behavior.
Product mismatch	The agent does not answer from another product’s source.
Prompt injection or secret request	Refusal or safe handoff without exposing prompts or credentials.
Provider failure	Operator-safe error or handoff state.
Human request	Clean operator handoff.

If a question fails, first check the Context tab for missing or unattached sources. Then tune behavior or guardrails only after the right source material is attached.

Agent Sections

Section	Use it for
Overview	Summary, status, widget key, context intelligence, and high-level controls.
Context	Source attachments and coverage.
Behavior	Persona, greeting, runbook instructions, and response style.
Guardrails	Safety rules, confidence, handoff behavior, and boundaries.

Agent Name And Color

Each agent has a display name and system color. Use these fields to make the agent easy to identify anywhere Woes shows routing or ownership.

Field	Where it appears	Guidance
Display name	Agents list, Inbox labels, conversation header, widget routing, analytics, and settings selectors.	Use the product, site, or support surface the agent owns.
System color	Agent labels, badges, and selector accents.	Pick a color that helps operators distinguish agents at a glance.

Changing an agent name or color does not change its widget key, attached sources, security settings, or conversation history. It only changes how the agent is presented across the workspace.

Default Agent

Every workspace should have a default agent. If the current default is deleted, Woes promotes the oldest remaining agent to default. Each agent can have its own widget public key. Use this when you want a different agent on each site, docs area, or product. Find all agent keys in Settings → Keys. When you install an agent widget key, that key controls the widget copy, attached context, identity verification settings, and routing for that site. If two sites need different behavior, give each one a different agent key.

Billing And Capacity

Agent capacity is plan-aware. The active plan includes a fixed number of agents and widgets, and extra agent/widget capacity can be added from Billing. If the workspace reaches its effective agent limit, Woes blocks new agent creation until an admin removes unused agents, upgrades the plan, or adds capacity.

Bring Your Own AI

Agent creation and source attachment are separate from model-provider billing. If the workspace uses Bring Your Own AI, saved agents keep the same grounding, confidence, clarification, and handoff rules. Only server-side model execution changes.

Woes documentation

Create and Manage Agents

Create and Manage Agents

When To Create More Than One Agent

Create an Agent

Test Before Routing Customers

Agent Sections

Agent Name And Color

Default Agent

Widget Keys

Billing And Capacity

Bring Your Own AI

​Create and Manage Agents

​When To Create More Than One Agent

​Create an Agent

​Test Before Routing Customers

​Agent Sections

​Agent Name And Color

​Default Agent

​Widget Keys

​Billing And Capacity

​Bring Your Own AI

Create and Manage Agents

When To Create More Than One Agent

Create an Agent

Test Before Routing Customers

Agent Sections

Agent Name And Color

Default Agent

Widget Keys

Billing And Capacity

Bring Your Own AI