General Overview

What are the primary purposes of Twain Agents?

Twain Agents serve two main functions:

Automation of Workflows: They create search agents to find "buying signals" and qualification agents to filter out accounts or contacts that don't fit your ideal customer profile.
Continuous Improvement: They are designed to get smarter over time. The more feedback you give them, the better they become at representing your brand.

How many agents do I need?

One agent per use case — not one per campaign. The main types are:

Outbound agent — for cold outreach campaigns
Inbound agent — for responding to inbound leads
CRM nurture agent — for warming existing contacts

If all your campaigns are outbound, one outbound agent handles all of them. When you create a new campaign, just select the same agent — no need to rebuild anything. You do not need a separate agent per campaign.

Setting Up Your First Agent

How do I start creating an agent?

Click "Create Agent" and enter your website domain.

Important: Your domain identifies the product the agent represents and cannot be changed later.

What information should I provide during setup?

Purpose & Audience: A "meta-prompt" explaining who you are targeting and why.
Collateral: You can upload files (like an lm.txt or company brochures) that describe your product and value proposition.
Buying Signals: Define what triggers an outreach (e.g., a specific company milestone or tech stack change) using natural language.

Can the agent help me define my strategy?

Yes. Once you provide your initial inputs, Twain will automatically propose:

Buyer Personas: Specific roles to target.
Signals & Filters: Automated research steps to identify and qualify leads.

Refining Agent Intelligence

How do I control the agent's writing style?

In the Writing Style settings, you can:

Set a specific Agent Tone.
Create a Block List for words or phrases. Twain blocks these at the token level, so you don't have to keep reminding the AI "never to use" them in every prompt.
Define Company Vocabulary and unique value propositions.
Write language-specific rules. For example: "Always use formal salutation in German (Sie, not du)." Rules like this are stored in the agent and apply to every campaign automatically.

How does the brainstorming/review feature work?

On the right side of the interface, the agent is "aware" of all your settings and has web access. You can click "Review my setup" to have the agent analyze your configuration and suggest improvements. You can then ask the agent to apply those changes automatically.

Signals & Clay Integration

How do I use signals from Clay in Twain?

If you're passing leads from Clay, map your signal columns from the Clay table as custom variables in Twain. These can be account-level signals (e.g., a funding round) or contact-level signals (e.g., a recent LinkedIn post). In Twain, they're all just referenced as "signal" — the label doesn't matter as long as the variable is mapped correctly from the Clay table.

To use a signal in your messaging, open the Framework Editor and add a conditional that references the signal variable. For example: if signal is present, reference it in the opener.

Does Twain's qualification replace what I do in Clay?

Run both — they complement each other. Twain's disqualification is more nuanced than what you can do in Clay. In addition to hard rules (minimum role tenure, company size, location), Twain catches soft disqualifications: contacts who've recently left a company, outdated signals, or other cues that suggest poor timing.

You'll see these flagged in the Warnings view inside your campaign. A lead with a warning isn't automatically excluded — you review and decide. Adding your own qualification rules in the agent does no harm; Twain applies both yours and what it finds independently.

How should I use signals in my sequences?

You have a few options — experiment to find what works best:

Use the mapped Clay signal in message 1, and let Twain find its own insight for message 2.
Let Twain find signals independently for all messages.
Reference the mapped signal in the opener, and weave in Twain's own research in follow-ups.

Using Agents in Campaigns

Do I need to rewrite my prompts for every campaign?

No. Because the agent already stores your research signals and qualification rules, your campaign descriptions can focus purely on intent and context. You don't need to repeat technical rules once they are saved to the agent.

Can I select specific personas for a campaign?

Yes. Even if an agent has five stored buyer personas, you can toggle them on or off for specific campaigns to ensure your messaging remains highly targeted.

The "Updates" Feature

What happens if I change something inside a specific campaign?

If you make a change while reviewing a campaign (e.g., adding a new blocked word), a blue dot will appear under the Updates tab.

What is the difference between a Local Change and a Source Agent update?

Local Change: The tweak only applies to that specific campaign.
Publish to Source Agent: By clicking this, you save the change to the "Master" Agent. This ensures that every future campaign benefits from that refinement, making the agent smarter the more you use it.