ChatBaser · Available now · Included in Pro

The AI agent builder grounded in your approved knowledge.

ChatBaser is an AI agent workspace for setting a focused job, attaching approved files and text, testing real conversations, and learning from the resulting chat history.

Product video · 1:51

Write the instructions. Add the source material. Test the agent in the playground. Then inspect its logs and revise the knowledge or behavior when an answer misses the mark.

Visit ChatBaser See the agent workflow

Current approved knowledge sources are file uploads and direct text. Widgets, public links, custom domains, integrations, and team workspaces are not promised here.

Inside ChatBaser

See ChatBaser in action.

Real screens from the workspace: test the agent in the playground, crawl the pages it learns from, and answer customers from a help center.

ChatBaser Playground with a branded green chat widget for a fictional outdoor store: the agent answers a Canada shipping question with delivery time and free-shipping threshold, then answers a used-tent return question with a 60-day policy and a cited source page, with copy, play, and thumbs up or down actions and suggested message chips

ChatBaser Website sources page after a finished crawl: a Crawl complete badge showing 34 pages found, a Link sources list grouped under one store domain with pages such as Shipping and Delivery Times, Returns and Used Gear Policy, and Tent Sizing Guide ticked, Preview Selected and Train Selected buttons, and a Sources panel showing 34 Links, 1.8 MB of 10 MB used, and Trained status

ChatBaser public Help Center portal in dark mode for a fictional outdoor store, with recent chats in the sidebar and an answered conversation: step-by-step instructions to change an order size, shipping timing, a link to the returns page, and a follow-up answer with live support hours and how to submit a ticket

What changes

Build the assistant that reads your business first.

Give every agent one clear job, approved knowledge, testable instructions, and a conversation record that shows exactly what needs to improve next.

GENERIC CHATAsk → hope → copy the answer → discover the gapThe business context lives outside the conversation.

CHATBASER AGENTDefine → ground → test → inspect → improveThe job, source material, and conversation record stay together.

Product facts

The product facts before you begin.

Status: Available now
Membership: Pro plan
Start with: Define and ground the agent
Review: Test real conversations
What moves forward: Learn from activity and refine

ChatBaser is available now in Pro. This page limits its knowledge sources and deployment claims to the capabilities verified in the current product.

How the work moves

From approved knowledge to an agent you can keep improving.

The AI answers from the instructions and sources you provide. You test the conversation, inspect the logs, and change the agent when the evidence says it needs work.

01INPUT

Define one job

Set the agent’s name, purpose, instructions, personality, identity, and approved files or direct text.

02AI WORK

Use the knowledge

ChatBaser processes the attached sources and makes them available during the agent’s conversation.

03OUTPUT

Run the conversation

Test multi-turn answers, retain chat history, and inspect activity through logs, charts, and exports.

04YOUR CONTROL

Correct the weak answer

Add or remove sources, revise the instructions, retrain, retest, and review the next conversation.

CREATEagent name and purpose

GROUNDfiles and direct text

TESTinteractive playground

LEARNlogs, charts, exports

01 · Give the assistant one clear job

Start with purpose, not a blank chat box.

Create the agent with a name and description, configure its system instructions and personality, and shape the colors, avatar, and position that identify it. The product keeps AI complexity behind the agent’s operating settings.

ChatBaser Agent Behavior settings with agent type cards (AI Agent, Support Agent, Sales Agent), a custom type 'Scale Plus Custom', and a Create New Custom Type button

PURPOSE

Name the job the agent owns.

Define whether it supports customers, answers product questions, helps a team, or guides a focused workflow.

ONE AGENT · ONE RESPONSIBILITY

INSTRUCTIONS

Set the behavior

Use the system prompt and personality configuration to establish the role, boundaries, and response style.

BEHAVIOR BEFORE ANSWERS

IDENTITY

Make it recognizable

Configure colors, avatars, and visual placement for the agent experience.

APPEARANCE CONTROLS

LIFECYCLE

Manage the agent

Create, open, configure, test, and retain multiple focused agents from the dashboard.

AGENT WORKSPACE

SAAS MODEL

No personal AI key

The platform manages the underlying AI connection rather than asking each user to configure an API key.

LESS TECHNICAL SETUP

BOUNDARY

Deploy features are evolving

Current marketing does not promise widget, iframe, public URL, custom domain, Slack, Discord, or webhook availability.

CURRENT PRODUCT ONLY

02 · Give it approved material

Ground the answer in the source the business controls.

Add files or direct text to the agent’s knowledge workspace and let the document-processing flow make those sources available during testing. Keep each agent’s sources focused on the job it owns.

File uploadsDirect textDocument processingPer-agent storageSource management

ChatBaser file sources list for a support assistant with markdown knowledge files such as user guides, briefs, and plan documents, each showing size and type

03 · Test before the customer does

Use real questions to find the weak answer.

Open the interactive playground, begin a conversation, review recent chats, and keep the history in database storage. The test surface makes it possible to refine the instructions or knowledge before wider deployment capabilities arrive.

CONVERSATIONStart, continue, or end a chatUse natural multi-turn questions to see whether the agent holds the context and follows its job.

HISTORYRecent chats + persistent recordsReturn to prior conversations and inspect how the answer developed across a thread.

REFINEChange the source or instructionTurn a weak answer into a specific correction to the agent’s knowledge or behavior.

04 · Questions are product research

Read the conversations as operating data.

Use date filtering, chat logs, details, responsive charts, pagination, and storage calculation to understand agent activity. Export logs in JSON, PDF, or CSV when the conversation history needs deeper review.

ChatBaser chat logs split view: conversation list on the left with a selected 'cant publish page' thread (12 messages), and the full transcript on the right where the agent replies with a step-by-step checklist

Chat analytics

Read the charts, not a single counter.

Choose the date range, then read actual conversation activity and where it came from across the agent collection.

ChatBaser analytics for a chat agent: 36 total conversations, 40 messages, contact capture metrics, a Chats started line chart rising sharply to Sep 28, and chats by country led by United States (14), United Kingdom (7), Canada and Germany (5)

LOGInspect the threadOpen the conversation and its details instead of reducing it to a single counter.

FILTERChoose the date rangeNarrow the activity to the period being reviewed.

MEASURERead the chartsUse actual database activity and storage usage across the agent collection.

EXPORTJSON · PDF · CSVTake the chat record into analysis, training, or support review workflows.

Feature Atlas · ChatBaser

See every ChatBaser feature—and what each one gets you.

Browse every ChatBaser feature, see exactly where it lives in the app, and find the one that moves your business forward today.

Explore the ChatBaser atlas 67 features across 11 areas

The useful assistant knows its job

Build from approved knowledge. Improve from actual questions.

Create the agent, ground it, test the conversation, and let the activity reveal what the business should explain better next.

Get Started, It’s Free Explore specialized GPT tools

Recently added

Let customers type, talk, or reach you anywhere.

ChatBaser added voice, a developer connection for every bot, and more control over what it knows.

New

Voice agents

Offer chat, voice, or both. Visitors tap the microphone, see the transcript, and hear the answer. Choose from 13 voices, set the speaking speed, and write a shorter greeting for voice.

New

Connect your bot to other tools

Every chatbot and help center is ready for voice agents, phone systems, support desks, automations, and custom apps. Create private keys, copy examples, or hand an AI coding agent everything with Copy Markdown for Agents.

New

Learns from your support team

Connect HDHelpDesk and your team's helpful replies become screened, reviewable knowledge for the chatbots you choose.

New

Easier knowledge upkeep

Import large sitemaps in batches, edit or delete saved text sources, copy answers in one click, and let visitors open a larger, resizable chat window.

Straight answers

What to know before you open ChatBaser.

What is ChatBaser?

ChatBaser is an AI-agent workspace for defining one job, grounding it in approved files or text, testing conversations, and reviewing activity and analytics.

What do I bring to ChatBaser?

Bring the agent instructions, identity, approved source files or direct text, and the boundaries for the job the agent should perform.

What does ChatBaser produce?

It produces a configured agent, persistent playground conversations, activity logs, analytics, storage reporting, and supported JSON, PDF, or CSV exports.

What remains under my control?

You control the instructions, knowledge, enabled behavior, test conversations, revisions, and whether an answer is appropriate to use.

What does ChatBaser not promise?

Public links, widgets, iframes, custom domains, platform integrations, and team workspaces remain in development and are not presented as available.