How to build an agent, put it to work and pay for it. Every answer has its own link, so you can share it.
LotsAgent turns your docs, your skills and the apps it should use into an AI agent. The agent answers questions, takes action in your apps, and checks with the person before anything important.
You can put it inside your product as a sidebar, give it its own web address, or run it in your workflows through the API, MCP and schedules. One agent can do all three.
Sign up at lotsagent.com. You get a 14-day trial with one agent and 2,000 credits ($2), and no card is needed.
The setup wizard opens and takes you through your first agent in five steps. You can change every setting later.
Founders who want an agent in their own product or business, and AI consultants who build agents for their clients.
Coming in November for consultants and teams: team members with roles, a separate workspace for each client, and handing an agent over to a client when the work is done.
The console is where you set up agents and take them into your product, your site or your workflows.
Each agent has its own tabs: Overview, Knowledge, Skills, Tools, Deploy, Look, Activity and Settings. Test opens a chat with the agent so you can try it before people see it.
The wizard saves as you go. If you leave, the agent waits as a draft, and its Overview shows Finish setting it up.
The agent's unique name on LotsAgent. It is used in its web address (your-handle.lotsagent.com) and in API paths.
If a handle is taken or reserved, the wizard says so and you pick another one.
New agents run on Claude Haiku 5.5, the default, or Claude Sonnet 5.5. Haiku is fast and costs the least. Sonnet is for harder work and costs more per answer.
You pick the model when you make the agent, and you can change it in the agent's Settings.
Websites, text you paste, and files. Add them in the wizard or on the agent's Knowledge tab.
The agent answers from your knowledge first, and says so when your knowledge does not cover a question instead of making something up.
On the Knowledge tab, add a website by its address. Choose a page and the pages under it, one page, or a sitemap.
LotsAgent reads up to 200 pages for each website, and an agent can have up to 20 websites. It follows your site's robots.txt. Pages appear in the list as they are read.
Each website updates automatically every day or every week, or only by hand. Pick the schedule next to the website, and click Re-sync now at any time.
A re-sync reads changed pages again and drops pages that are gone, so answers follow your docs.
Yes. When an answer comes from your knowledge, it links the page it came from: the page on your site for an imported website, or the public page you gave for pasted text.
When your help has a link for each topic, the answer links that topic, not just the page.
Text, Markdown, CSV and PDF files, up to 25 MB each.
When people ask something your knowledge does not answer, the agent notes the topic. The Knowledge tab lists these questions, most asked first, so you know what to write next.
Add help for a question, then mark it as covered. Only the topic of each question is kept, in the agent's own words, never the conversation. Each one is kept for 90 days after it was last asked.
A skill teaches an agent how to do a job your way, such as how to answer a refund request or how to write a weekly report.
Write your own, or import one from GitHub, and add it to any of your agents on the Skills tab.
On the agent's Tools tab, connect apps like Notion, Slack and Google Drive, or your own app through MCP. Then choose which actions the agent may use.
Anything that sends, deletes or moves money waits for the person to click Confirm.
Usually each person connects their own account. On a private agent you can instead connect an app for the agent itself, shared by everyone who uses it: on the Tools tab, choose Connect for this agent.
If someone else owns the account, send them an Invite someone to connect link. They connect it without a LotsAgent account. The link lasts 7 days, works once for each app, and you can revoke it.
Shared connections work only on private agents.
On the agent's Deploy tab, open Web address. Turn on its address at your-handle.lotsagent.com, or add your own domain.
People open the address and chat with the agent full screen. A help-center agent that answers from your docs lets anyone chat, with no sign-in.
On Deploy, Web address, enter your domain, for example help.yourcompany.com. Add the DNS record it shows at your domain provider, then click Verify. The address works once it is verified.
Every plan includes one custom domain for each agent.
Yes. Set up sign-in for the sidebar first. Then, on Deploy, Web address, turn on People sign in through your app on the web address.
People who are signed in to your app open the address and chat as themselves, full width, with the same rights as in your sidebar.
Yes, when the agent also has its own web address. The sidebar header shows Open this chat in a new tab, and the same chat continues full width there.
On the agent's Deploy tab, open Tasks and schedules. Save instructions the agent runs by hand, once at a set time, or on a schedule, such as every weekday at 9.
Every run shows in Activity with what it did and what it cost. A scheduled run has nobody to press Confirm, so an action that needs Confirm stops the run, and the run tells you which action waits for your approval.
Make an API key in Settings, API Keys, and give it Chat access. Then send a message to https://lotsagent.com/api/v1/agents/your-handle/chat with the key as a Bearer token.
You can stream the answer, continue a chat with its session_id, or start a background run and check on it later. The API docs have every request and answer.
On Deploy, API and MCP, turn on Allow MCP connections. It is off by default. Then add the agent's MCP address to your client, such as Claude, ChatGPT or Cursor.
The client signs in by itself, or you can send an API key.
You get 14 days with one agent and 2,000 credits ($2). No card is needed.
When the trial credits are used up or the 14 days end, pick a plan to keep your agents running. Nothing is deleted.
Plans are priced by the number of agents, and every plan includes credits every month.
Yearly billing gives you 2 months free. Every agent includes its own knowledge base, all three ways to deploy, one custom domain and unlimited end users. See the pricing page for the details.
Credits pay for the AI and the work your agents do: answering, searching your docs and using your tools. 1,000 credits are $1.
You pay what the AI costs, plus 20%. Your plan's included credits and the credits you buy are spent at the same rate. Unused included credits carry over, up to three months' worth. Buy more at any time from $10. Credits you buy never expire.
Your agents keep working on credits you buy. With no credits left, they stop answering until you add more.
At the monthly limit, people see: This assistant has reached its monthly limit. It will be back next month.
Yes, at any time from Billing. Moving up takes effect at once. Moving down or cancelling takes effect at the end of your billing period.
Nothing is deleted. If you have more agents than your plan includes, the agents you made first keep running and the rest pause until you move up or archive one.
No.
For the sidebar, the agent's Activity tab shows conversations, people, messages, tool calls, errors, Confirm cards and thumbs, per day and as a timeline for each conversation. It does not show what people wrote.
Your own test chats are yours to read on the Test screen.
You choose, from 1 to 365 days. 30 days by default, then they are deleted.
No. The chat runs in its own secure frame. Your page cannot read it, and the agent cannot read your page.
People see your notice, in your words, before their first message.
Use the feedback button in the console, or email contact@lotstech.com with the agent's handle and what you saw. Developers: the embed docs and the API docs have every setting, code example and error code.
How to build an agent, put it to work and pay for it. Every answer has its own link, so you can share it.
LotsAgent turns your docs, your skills and the apps it should use into an AI agent. The agent answers questions, takes action in your apps, and checks with the person before anything important.
You can put it inside your product as a sidebar, give it its own web address, or run it in your workflows through the API, MCP and schedules. One agent can do all three.
Sign up at lotsagent.com. You get a 14-day trial with one agent and 2,000 credits ($2), and no card is needed.
The setup wizard opens and takes you through your first agent in five steps. You can change every setting later.
Founders who want an agent in their own product or business, and AI consultants who build agents for their clients.
Coming in November for consultants and teams: team members with roles, a separate workspace for each client, and handing an agent over to a client when the work is done.
The console is where you set up agents and take them into your product, your site or your workflows.
Each agent has its own tabs: Overview, Knowledge, Skills, Tools, Deploy, Look, Activity and Settings. Test opens a chat with the agent so you can try it before people see it.
The wizard saves as you go. If you leave, the agent waits as a draft, and its Overview shows Finish setting it up.
The agent's unique name on LotsAgent. It is used in its web address (your-handle.lotsagent.com) and in API paths.
If a handle is taken or reserved, the wizard says so and you pick another one.
New agents run on Claude Haiku 5.5, the default, or Claude Sonnet 5.5. Haiku is fast and costs the least. Sonnet is for harder work and costs more per answer.
You pick the model when you make the agent, and you can change it in the agent's Settings.
Websites, text you paste, and files. Add them in the wizard or on the agent's Knowledge tab.
The agent answers from your knowledge first, and says so when your knowledge does not cover a question instead of making something up.
On the Knowledge tab, add a website by its address. Choose a page and the pages under it, one page, or a sitemap.
LotsAgent reads up to 200 pages for each website, and an agent can have up to 20 websites. It follows your site's robots.txt. Pages appear in the list as they are read.
Each website updates automatically every day or every week, or only by hand. Pick the schedule next to the website, and click Re-sync now at any time.
A re-sync reads changed pages again and drops pages that are gone, so answers follow your docs.
Yes. When an answer comes from your knowledge, it links the page it came from: the page on your site for an imported website, or the public page you gave for pasted text.
When your help has a link for each topic, the answer links that topic, not just the page.
Text, Markdown, CSV and PDF files, up to 25 MB each.
When people ask something your knowledge does not answer, the agent notes the topic. The Knowledge tab lists these questions, most asked first, so you know what to write next.
Add help for a question, then mark it as covered. Only the topic of each question is kept, in the agent's own words, never the conversation. Each one is kept for 90 days after it was last asked.
A skill teaches an agent how to do a job your way, such as how to answer a refund request or how to write a weekly report.
Write your own, or import one from GitHub, and add it to any of your agents on the Skills tab.
On the agent's Tools tab, connect apps like Notion, Slack and Google Drive, or your own app through MCP. Then choose which actions the agent may use.
Anything that sends, deletes or moves money waits for the person to click Confirm.
Usually each person connects their own account. On a private agent you can instead connect an app for the agent itself, shared by everyone who uses it: on the Tools tab, choose Connect for this agent.
If someone else owns the account, send them an Invite someone to connect link. They connect it without a LotsAgent account. The link lasts 7 days, works once for each app, and you can revoke it.
Shared connections work only on private agents.
On the agent's Deploy tab, open Web address. Turn on its address at your-handle.lotsagent.com, or add your own domain.
People open the address and chat with the agent full screen. A help-center agent that answers from your docs lets anyone chat, with no sign-in.
On Deploy, Web address, enter your domain, for example help.yourcompany.com. Add the DNS record it shows at your domain provider, then click Verify. The address works once it is verified.
Every plan includes one custom domain for each agent.
Yes. Set up sign-in for the sidebar first. Then, on Deploy, Web address, turn on People sign in through your app on the web address.
People who are signed in to your app open the address and chat as themselves, full width, with the same rights as in your sidebar.
Yes, when the agent also has its own web address. The sidebar header shows Open this chat in a new tab, and the same chat continues full width there.
On the agent's Deploy tab, open Tasks and schedules. Save instructions the agent runs by hand, once at a set time, or on a schedule, such as every weekday at 9.
Every run shows in Activity with what it did and what it cost. A scheduled run has nobody to press Confirm, so an action that needs Confirm stops the run, and the run tells you which action waits for your approval.
Make an API key in Settings, API Keys, and give it Chat access. Then send a message to https://lotsagent.com/api/v1/agents/your-handle/chat with the key as a Bearer token.
You can stream the answer, continue a chat with its session_id, or start a background run and check on it later. The API docs have every request and answer.
On Deploy, API and MCP, turn on Allow MCP connections. It is off by default. Then add the agent's MCP address to your client, such as Claude, ChatGPT or Cursor.
The client signs in by itself, or you can send an API key.
You get 14 days with one agent and 2,000 credits ($2). No card is needed.
When the trial credits are used up or the 14 days end, pick a plan to keep your agents running. Nothing is deleted.
Plans are priced by the number of agents, and every plan includes credits every month.
Yearly billing gives you 2 months free. Every agent includes its own knowledge base, all three ways to deploy, one custom domain and unlimited end users. See the pricing page for the details.
Credits pay for the AI and the work your agents do: answering, searching your docs and using your tools. 1,000 credits are $1.
You pay what the AI costs, plus 20%. Your plan's included credits and the credits you buy are spent at the same rate. Unused included credits carry over, up to three months' worth. Buy more at any time from $10. Credits you buy never expire.
Your agents keep working on credits you buy. With no credits left, they stop answering until you add more.
At the monthly limit, people see: This assistant has reached its monthly limit. It will be back next month.
Yes, at any time from Billing. Moving up takes effect at once. Moving down or cancelling takes effect at the end of your billing period.
Nothing is deleted. If you have more agents than your plan includes, the agents you made first keep running and the rest pause until you move up or archive one.
No.
For the sidebar, the agent's Activity tab shows conversations, people, messages, tool calls, errors, Confirm cards and thumbs, per day and as a timeline for each conversation. It does not show what people wrote.
Your own test chats are yours to read on the Test screen.
You choose, from 1 to 365 days. 30 days by default, then they are deleted.
No. The chat runs in its own secure frame. Your page cannot read it, and the agent cannot read your page.
People see your notice, in your words, before their first message.
Use the feedback button in the console, or email contact@lotstech.com with the agent's handle and what you saw. Developers: the embed docs and the API docs have every setting, code example and error code.
