Chatsonic
AISend prompts to Chatsonic, Writesonic's assistant, and receive its reply with the supporting context. Agents draft and rewrite marketing copy inside a workflow that already has the customer record.
What This Integration Enables
Chatsonic is Writesonic's assistant, and the reason to reach for it inside a flow is not that it writes. Plenty of things write. It is Use Google Results, which grounds the answer in live search rather than model recall, so the assistant can respond about something that happened last week. That makes it a research step rather than a drafting step, and treating it as one is the difference between a useful integration and an expensive paraphrase of a model you already have connected.
The connector is a single action, and its two design details shape everything built on it. The prompt is limited to 600 characters, which forces the question to be specific and, usefully, keeps every prompt short enough for a human reviewer to read in full. And Chatsonic is stateless: to hold a multi-turn exchange you enable Use Memory and pass the previous turns back in Conversation History, flagging each entry with Is Sent set to true for a user prompt and false for a Chatsonic reply. The memory lives in your flow, not in the vendor's session store. That is more work and considerably more control, because the conversation state is inspectable, editable and auditable in the place where the rest of the workflow already is. Answers come back in any of twenty five supported languages, with any generated image URLs alongside.
Without FlowRunner
With FlowRunner
Use Case Scenarios
A current opening line, checked before it is sent
An account enters the sequence in Apollo or Instantly. The agent calls Chat With Chatsonic with Use Google Results enabled and a tight prompt: the company name, and the specific question worth asking about them this quarter. The reply is written onto the account record in HubSpot together with the prompt that produced it, so the rep sees the claim and the question side by side. The rep confirms it, or corrects it, and the message goes out with an opener that was true this morning.
An inbound question answered with something current
A form submission or support message asks about something recent: a regulation, a competitor announcement, a pricing change. The agent poses the question to Chatsonic with grounding on, gets a current answer, and drafts a response for the owning team in Slack with the answer and the original question attached. For anything that needs a follow-up question, the flow enables Use Memory and replays the prior turns in Conversation History, so the second question builds on the first rather than starting cold.
The same message, twenty five ways
A campaign needs localised versions. The agent takes the approved English message and asks Chatsonic for each target language in turn, holding the approved source constant so every version derives from the same reviewed text rather than from a chain of translations. Because the prompt cap is 600 characters, long copy is localised section by section, which is a constraint that happens to produce better review granularity. A native speaker signs off on each language before it enters the sending tool.
Human-in-Loop Highlight
Google grounding is what makes this connector worth having, and it is also exactly what the gate is for. Chatsonic returns a confident answer and no citation payload, so a flow gets the claim without the sources behind it. Most of the time the claim is right. The failure that matters is narrow and expensive: a company with a common name, an acquisition attributed to the wrong entity, an article from three years ago read as current. A wrong fact in an internal summary is an annoyance. The same wrong fact in a cold email is a message that tells the recipient, in their own inbox, that you did not actually look. That does not get retracted. So the agent stops before the send and puts both halves in front of the rep, in a Flow conversation they can answer without leaving what they are doing: "For Meridian Logistics I asked: what did they announce in the last six months? The answer is a Series B in April. There are two companies with this name and the account domain does not match the one in the result. Use it, replace it, or send without an opener?" One reply, and the flow either continues or drops the line. That is human-in-the-loop sized to the actual risk: not a review queue for every message, just a stop at the one place where being confidently wrong reaches a customer.
Agent Capabilities
1 actionsChat
1- Chat With Chatsonic Sends a prompt to Chatsonic and returns its reply together with any images it produced. With Use Google Results enabled the assistant grounds its answer in live Google search results, which is what lets it respond about recent events rather than model knowledge alone. Chatsonic is stateless, so multi-turn exchanges enable Use Memory and pass the previous turns in Conversation History, flagging each entry with Is Sent set to true for a user prompt and false for a Chatsonic reply. The prompt is limited to 600 characters, and the assistant answers in any of 25 supported languages.
Frequently Asked Questions
What can FlowRunner do with Chatsonic?
FlowRunner agents can run Chat With Chatsonic in Chatsonic.
Does connecting Chatsonic to FlowRunner require OAuth?
No. Chatsonic connects to FlowRunner with an API key, no OAuth flow required.
Can Chatsonic trigger a FlowRunner workflow automatically?
Chatsonic doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with Chatsonic
$100 in credits. No card required. Connect in minutes.