What is Agentency?

What it is

Agentency is a platform for building chatbots that answer from your own material — your PDFs, your policies, your website — instead of guessing from the open web. You upload what you know, Agentency turns it into something the chatbot can search, and visitors get answers grounded in that material with the sources attached.

A chatbot in Agentency is one self-contained thing. It owns its own knowledge, its own instructions, its own look, its own conversations, and its own places to appear. Two chatbots in the same account never share knowledge or conversations. That is deliberate: a support chatbot trained on your refund policy should not be answering pre-sales questions out of a pricing deck it was never given.

Three things make it different from a general-purpose chat tool:

• It is grounded. If the answer is not in your material, the chatbot says so instead of inventing one. You control the wording of that "I don't know" reply.
• It can act, not only talk. Call actions let a chatbot send an email, post to your systems over HTTP, or notify your team — but only the actions you explicitly build and switch on.
• It is multi-tenant and role-aware. You own an account, you invite teammates into it with limited roles, and they never get your billing or your API keys unless you are the owner.

When you would use it

Read this page when you are deciding whether Agentency fits, or when you want a map of the product before clicking around. Typical fits:

• A support team drowning in the same twenty questions, with the answers already written down somewhere nobody reads.
• A sales site that needs to qualify visitors and capture a name and email before a human gets involved.
• An internal handbook, HR policy set, or product manual that staff cannot navigate.
• A storefront where the shipping, returns, and sizing answers all live in different documents.

It is not a ticket queue and not a live-chat product with human operators taking turns. Conversations are stored so you can read and review them, but there is no operator console where a person types back as the chatbot. If a question needs a human, you build a handoff that hands off to whatever your team already uses.

Where to find it

Open the marketing home page, then Dashboard after you sign in.

The five pieces you will actually use

1. The account

The account is your workspace and it belongs to one person — the owner. Everything (chatbots, credits, billing, knowledge) hangs off that account. Teammates are invited into your account rather than getting a copy of it. If you belong to more than one account, you switch between them from Accounts in the sidebar, and the whole dashboard re-scopes. See What is an account.

2. The chatbot

Created from Dashboard → Chatbots → Create, optionally from a starter template. A chatbot has two independent states that beginners constantly confuse:

• Activation — draft, active, or inactive. Decides whether anyone outside your dashboard can reach it at all.
• Training — whether its knowledge has finished being processed. Decides whether it can answer well.

They do not imply each other. Chatbot readiness explains both lights in one page and is worth reading before you go live.

3. Knowledge

Files you upload, text you paste, cloud documents you import, and websites you crawl. Agentency processes each source into searchable pieces; while that is running the source shows as training, and when it finishes the chatbot can cite it. Start at Knowledge overview.

4. Instructions and Call actions

Instructions are short rules that shape every reply — tone, what to never say, what to always mention. Call actions are the things the chatbot may do: send a notification, hit an HTTP endpoint you configure, hand a visitor to a person. See Write instructions and What are Call actions.

5. Channels

Where visitors reach the chatbot:

• Web widget — a chat bubble on your own site, from a <script> tag. See Embed the website widget.
• Direct link — a hosted page at /agent/{handle} you can send to anyone. See Share a hosted link.
• Iframe — the same hosted page inside a frame on a site builder.
• Messaging apps and plugins — availability varies by deployment and plan.

Steps

1. Open the Agentency website and read Features, How it works, and Pricing if you have not signed up yet.
2. Create a free account from Start free. You land on Dashboard. See Create your account.
3. Open Chatbots in the sidebar and create your first one — Create your first chatbot.
4. Open the chatbot to see its seven tabs: Playground, Conversations, Analytics, Knowledge, Actions, Integrations, Settings.
5. Add knowledge, wait for training to finish, then ask it something with the Test Chatbot button in the chatbot header — Try Test Chatbot.
6. When the answers look right, work through the Go-live checklist.

What you will see

The sidebar shows Dashboard, Chatbots, Customers, Accounts, and Settings, with a credits meter. The top bar has a search you can open with Ctrl/Cmd+K, a help entry, notifications, and your user menu. Settings has its own children — Account, Security, Members, Billing, Appearance, Developer — and Billing and Developer only appear for the account owner.

Inside one chatbot, the seven tabs sit under the chatbot's name, and a Test Chatbot button sits in the header on every one of them.

Limits and plan notes

The Free plan gives you 50 message credits, 1 chatbot, 400 KB of knowledge, 1 custom instruction, no Call actions, one seat (you), and no integrations — meaning no website widget, no plugins, and no messaging channels. You can still activate the chatbot and share its hosted /agent/{handle} link, which is enough to try the product end to end for real.

Paid plans raise every cap and switch integrations on. Plans and what you get has the full table; Message credits explains what actually consumes a credit.

Common problems

I signed in and the dashboard looks empty.

Normal on a new account — there is nothing to chart yet. Create a chatbot from Chatbots → Create. The dashboard fills in once conversations start.

I cannot find a feature I saw on the marketing site.

Some surfaces are deployment-dependent, particularly social messaging channels. If a channel shows as unavailable or coming soon, that is a platform-level state, not something wrong with your account. See Messaging channels availability.

Where do I choose which AI model answers?

Every chatbot starts on the default Agentency AI model, at 1X — one credit per reply. The Free plan always uses it. On a paid plan, open the chatbot's Playground tab and select Change AI model on the AI model card; each model shows its credit multiplier, from 1X to 12X. See Message credits and Chatbot AI settings.