AI Agent · Claude
A Claude-powered support agent that stays on policy
Run your ChatFlo agent on Anthropic's Claude models. Claude is unusually good at following a written policy closely and holding a consistent tone across a long conversation — which is most of what support writing actually is.
Can ChatFlo run my support agent on Claude?
ChatFlo can run your support agent on Anthropic's Claude models — Sonnet 5, Opus 4.8, or Haiku 4.5 — grounded in your own documentation. No Anthropic account or API key is required; usage is included in your plan. Claude is the usual pick when replies must follow a written refund, shipping, or eligibility policy closely.
Why Claude
The one to pick when tone and policy accuracy matter most.
Follows written policy closely
If your refund window, eligibility rules, or shipping exceptions are written down, Claude tends to apply them as written rather than paraphrasing them into something looser.
Consistent tone over a long thread
Brand voice holds up across a fifteen-message conversation, not just the opening reply — which is where most support automation starts to sound generic.
Careful about what it does not know
Claude is comparatively willing to say a question needs a person. On a support queue that is a feature: a clean escalation beats a confident wrong answer.
Available models
The Anthropic models you can select
Pick one in AI settings. Switching is a setting, not a migration — your training content, guardrails, and history stay where they are.
Claude Sonnet 5
Smart and balanced — the usual pick
Claude Opus 4.8
Most capable for complex threads
Claude Haiku 4.5
Fast and cheap for high volume
When another model is the better call
For very high-volume, low-complexity FAQ traffic, GPT-4o mini or Haiku 4.5 will cost less per conversation with little practical difference in answer quality.
Run your agent on Claude
Switch model in AI settings — your training content, guardrails, and history stay exactly as they are.
Frequently asked questions
Mostly when your answers have to track a written policy precisely — refund windows, eligibility criteria, regulated wording — or when a consistent brand voice across a long conversation matters more than raw speed. For general FAQ deflection the two are close, and you can switch between them without retraining anything.
Sonnet 5 for almost everything. Move to Opus 4.8 if your conversations involve multi-step reasoning over several documents, and to Haiku 4.5 if you are handling very high volumes of routine questions and want to cut cost per conversation.
No. ChatFlo runs the model for you on every paid plan — there is no separate Anthropic account, billing relationship, or key to manage. Model usage is included in your plan.
Yes, from the AI settings screen, at any time. The model is a setting on your agent, not a plan tier — your training content, guardrails, tone, and conversation history are untouched by the switch, so you can A/B two models on the same setup.
No. Training is the same regardless of model: you point ChatFlo at your website, docs, and files, and it indexes them. The model only affects how answers are composed from that material — it never changes what the agent is allowed to draw on.
Other models
ChatFlo is not locked to one model vendor
Your agent runs on whichever of these you select, and you can change your mind without rebuilding anything.