Chatbase
Also known as: Chatbase.co
No-code builder for support agents grounded in ingested docs, help center articles and support tickets, with AI Actions that manage orders, tickets and scheduling in connected systems, deployed across six native channels plus a REST API.
Chatbase is a no-code platform for building and deploying custom AI support agents trained on a company's own data. Founded in 2023 and bootstrapped to serve thousands of businesses, it was one of the early chat-with-your-data products, and its pitch is letting a non-technical team stand up a ChatGPT-style agent for their business in minutes, without building a custom stack.
The starting point is your knowledge. You feed Chatbase website URLs, which it crawls, along with PDFs, documents, spreadsheets, raw text, Q&A pairs, or a Notion workspace, and it can even ingest existing Zendesk and Salesforce support tickets as training data. From that material it builds a conversational agent that answers customer questions grounded in your content.
What pushes Chatbase past a simple FAQ bot is AI Actions, which let the agent take real steps in external systems. Connected to tools like order management, a CRM, or a helpdesk, an agent can look up an order, update a subscription, change a customer's address, capture a lead, or book a meeting, with integrations including Stripe, Shopify, Calendly, Slack, Zendesk, and Salesforce. When a query needs a person, the agent escalates to a human via live chat or a helpdesk ticket based on instructions you write in plain language.
A distinctive strength is model flexibility. Chatbase offers many leading models from several providers, including OpenAI, Anthropic, Google, and xAI, and lets you switch models or compare them side by side in a Playground where you also set the system prompt, configure actions, and test responses live.
Agents deploy across a website widget, WhatsApp, Messenger, Instagram, Slack, and email, with chat localization across many languages. Chatbase adds analytics on resolutions and engagement so the agent improves over time, a public API for deeper integration, and enterprise controls like SSO, encryption, and compliance, fitting teams that want a hosted product instead of a custom runtime they maintain.
Vendor details
Canonical URL
https://chatbase.co
Category
Customer support agent
Subcategory
Support — chatbot builder
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
In practice
You want a support agent live this week, not a six-month engineering project. Chatbase trains on your website, PDFs, and help docs, and you deploy a working agent on your site, WhatsApp, or Slack in minutes.
Your bot can answer questions but can't actually do anything. Chatbase's AI Actions connect it to your order system and CRM, so it looks up orders, updates subscriptions, and changes addresses, not just replies.
You don't want to bet your support on one AI provider. Chatbase lets you pick from models across OpenAI, Anthropic, Google, and xAI, and compare them side by side in the Playground before you commit.
Agentic Index coverage score
10.0 / 14 capabilities · 71%
| Integrations & Tool Calling | Full |
|---|---|
|
AI Actions ship with native connectors to Stripe, Shopify, Zendesk, Salesforce, Calendly, Cal.com and Slack, with WhatsApp and Messenger alongside, and they can write as well as read. A Custom Action endpoint accepts any REST API, so coverage is not limited to a fixed catalog. Real time connectors sync to CRMs, order systems and help desks, so replies reflect live business records. The model decides when to trigger an action from conversation context, with guardrails preventing unintended operations. Sourcechatbase.co/blog/chatbase-vs-custom-chatbotread 2026-09-05 |
|
| Workflow Orchestration | Full |
|
AI Actions carry out transactional work end to end. They manage orders through Shopify, pull subscription and invoice details from Stripe for verified users, create and auto assign Zendesk tickets with drafted responses, schedule through Calendly and Cal.com, send Slack alerts and update records. The model decides when to trigger an action from conversation context, and guardrails prevent unintended operations. Escalation is part of the same chain, handing to a live agent through Salesforce Omni Channel with real time presence and full conversation context carried across. Identity and Contacts authorize actions for known users, so a transaction is bound to a verified customer. Sourcechatbase.co/blog/the-ai-agent-playbookread 2026-09-05 |
|
| Knowledge Grounding & RAG | Full |
|
Chatbase ingests documents, web pages, help center articles and raw text into a searchable knowledge base, retrieved at query time through retrieval augmented generation. It can also ingest Zendesk and Salesforce support tickets as training data. Source Suggestions identify knowledge gaps from live conversations automatically, and auto retraining keeps the index in sync as source content changes. AI powered guardrails ground responses in the configured sources. Sourcechatbase.co/blog/the-ai-agent-playbookread 2026-09-05 |
|
| Human Oversight & Guardrails | Full |
|
A proprietary prompt architecture prevents unconstrained generation by holding the model inside the configured knowledge scope, so the agent refuses to answer outside its domain. Guardrails prevent unintended operations when the model decides to trigger an action. Operators can correct responses directly from chat logs with the Improve Answer feature, and the correction persists. Live agent handoff carries full context through Salesforce Omni Channel. Role based access uses custom roles across fourteen permission areas, covering agents, sources, chat logs, contacts, integrations, billing, subscription, members, webhooks, workspace, API keys, actions, analytics and leads, and Enterprise adds SSO. Chatbase says no guardrail system is perfect and that hallucination prevention is probabilistic. Sourcechatbase.co/blog/chatbase-vs-custom-chatbotread 2026-09-05 |
|
| Security, Identity & Governance | Full |
|
Chatbase is SOC 2 Type II and GDPR compliant, with AES 256 encryption at rest and in transit. The Enterprise plan includes SSO, and RBAC uses custom roles across 14 permission areas. Per user rate limiting applies, domain allowlisting controls where the agent can be embedded, and customer data is never used to train models. A trust center at trust.chatbase.co lists the subprocessors, and the DPA commits to a subscribe mechanism that tells customers about new subprocessors. Sourcetrust.chatbase.co/subprocessorsread 2026-09-05 |
|
| Observability & Auditability | Partial |
|
Backstage is a workspace to debug, inspect and improve an agent. Analytics cover conversation logs, sentiment, resolution metrics and dashboards, with topic clustering, confidence scoring and geographic distribution, export to JSON, PDF and CSV, and reviewable chat logs behind them. There is no decision trace for each step, tool call log or retrieval record for an individual run, and no configuration audit trail. A team can read what was said and how the agent performed, but cannot reconstruct which action fired and why. Sourcechatbase.co/llm-inforead 2026-09-05 |
|
| Memory & State Persistence | Partial |
|
Identity and Contacts personalize and authorize actions for known users, holding customer records alongside persisted chat logs. Full conversation context transfers when a conversation hands to a live agent. There is no agent memory across sessions with a set scope or lifetime. What persists is a contact record and a transcript. Sourcechatbase.co/blog/the-ai-agent-playbookread 2026-09-05 |
|
| Deployment & Data Residency | Not documented |
|
Hosting is fixed in the United States. Customer personal information is stored in databases hosted by third parties in the United States, with no other region offered. The DPA covers restricted transfers through Standard Contractual Clauses and points to a subprocessor list, with no residency option. There is no VPC, single tenant or on premises option. Sourcechatbase.co/legal/privacyread 2026-09-05 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Prebuilt AI Actions ship for common support and commerce jobs, covering order management, ticket creation, subscription lookup, scheduling and lead capture, and an existing agent can be duplicated as a starting point for a new one. There is no agent template library, marketplace or named catalog of prebuilt agents to browse and put into service. Customers adopt actions, not whole agents. Sourcechatbase.co/blog/the-ai-agent-playbookread 2026-09-05 |
|
| Triggers & Channel Coverage | Full |
|
One agent configuration reaches customers across six named channels and the phone. The channels are a customizable website widget or standalone hosted Agent Page, email, where the agent reads, drafts and responds to incoming support mail around the clock, WhatsApp, Facebook Messenger, Instagram and Slack, plus telephony that connects the agent to phone lines to handle inbound calls with natural voice. Full API access is available for custom surfaces, and the chat interface is localized into more than forty languages, including right to left support. Sourcechatbase.co/llm-inforead 2026-09-05 |
|
| Model Flexibility & Routing | Full |
|
The customer chooses the model. The no code builder lets an operator set the AI model alongside system instructions when creating an agent, and an operator can compare models directly for the same agent. Sourcechatbase.co/blog/chatbase-vs-custom-chatbotread 2026-09-05 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Outside systems can call the platform. REST API v1, and v2 in beta, give programmatic access to agent management, conversations and data, covering chatting with agents, creating and configuring them, and retrieving conversations, leads and analytics. A JavaScript embed supports identity verification through JWT and HMAC for authenticated sessions. Full API access is listed as a deployment surface in its own right, alongside the native channels, and what is available varies by pricing tier. Sourcechatbase.co/blog/chatbase-vs-custom-chatbotread 2026-09-05 |
|
| Testing, Debugging & Optimization | Partial |
|
Testing is named as a platform capability that validates agent behavior before and after deployment, and Backstage, described as the agent's control room, is a workspace to debug, inspect and improve it. Confidence scoring flags weak answers in production, and the Improve Answer feature lets operators correct responses from chat logs so the correction persists. There is no scored evaluation, scenario set, regression suite or comparable verdict on a proposed change, so a team can exercise an agent but cannot compare two versions of it. Chatbase's page for AI assistants names Testing and Backstage without saying how they work. Sourcechatbase.co/llm-inforead 2026-09-05 |
|
| Browser & Computer Use | Not documented |
|
No browser control, hosted session or computer use appears on Chatbase's product surface. The agent acts through configured AI Actions against native connectors and a Custom Action endpoint for any REST API, and reaches customers through its named channels and the public API, not by operating an interface. Sourcechatbase.co/llm-inforead 2026-09-05 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Drafts and version history are now on all Chatbase accounts. Saved edits go into a draft while customers keep talking to the live agent, and each publish is stored as a version with highlighted diffs and restore at any time. Backstage, the built in operations assistant, can also work on a draft and restore earlier versions with the user's approval.
Bears on: Human approval / guardrails
View sourceChatbase now runs automated test conversations against an AI agent to check it behaves as expected after changes to instructions, model, data sources, procedures or actions. Tests can verify the agent called exactly the expected actions, return mocked data or errors, and run supported read only actions such as Shopify product lookups live, while write and custom actions cannot run live.
Bears on: Human approval / guardrails
View sourceChatbase expanded Shopify actions so agents can create orders from WhatsApp, Instagram, Facebook Messenger, Chat Bubble and Help Page conversations, using payment links, cash on delivery or free orders permitted by configured procedures. Agents can prepare exchanges as new orders, with a human approving or rejecting the exchange in Chatbase Helpdesk before it reaches Shopify. The same release adds product search from customer images and order tagging for damaged or missing items and their free replacements.
Bears on: Human approval / guardrails
View sourcePricing
From $40 a month on Hobby, or $32 a month billed annually. A free tier is available.
Message credits. Each plan includes a monthly credit allowance, and every AI response draws credits weighted by the model used.
Included quota
Hobby, the entry paid plan at $40 a month or $32 billed annually, includes 700 message credits a month, 1 agent with advanced models, 5 AI Actions, 10MB of training data, 2 seats, API access, integrations and basic analytics. Free includes 50 credits a month, 1 agent and 1 MB of training data, with a watermark and no API or Actions, and free agents are deleted after 14 days idle.
What is public
Chatbase, an AI chatbot builder with over 10K paying customers, meters credits across five tiers, with 20% off for annual billing. Free is $0 with 50 credits a month and 1 agent, deleted after 14 days idle. Hobby is $32 a month annually or $40 monthly with 700 credits, 5 AI Actions and API access. Standard is $120 or $150 with 4,000 credits, 2 agents, voice and outbound. Pro is $400 or $500 with 15,000 credits, 5 agents and advanced analytics. Enterprise is custom, with an SLA, a CSM and white label. Earlier credit allotments were larger, with Hobby at 1,500, Standard 10,000 and Pro 40,000.
Billing mechanics
Every AI response consumes message credits, weighted by model. Standard models use 1 credit, GPT-5.2 and Gemini Pro use 2, Claude Sonnet 4.6 and Grok 3 use 3, Grok 4 uses 4 and Claude Opus 4.6 uses 5. A 700 credit Hobby plan is about 700 responses on cheap models but only about 140 on Opus. Credits reset monthly on the 1st, and when they run out the widget shows 'unavailable' unless auto recharge is on. Tiers also gate agents, AI Actions, training data size and seats.
Cost watchouts
Premium models burn credits 3 to 5 times faster than standard ones, and auto recharge can inflate bills when volume is erratic. Branding removal costs $39 a month and up, about $99 to $199 depending on tier. A custom domain is an add on at about $59 a month or comes with Enterprise, and extra agents are $7 a month each. Free agents are deleted after 14 days idle.
Variable cost rationale
Chatbase meters message credits. Every AI response draws credits, and premium models draw up to 5 per response, so the bill grows with conversation volume and model choice. A traffic spike or a switch to a premium model can multiply spend fast. By default the monthly allowance is a hard stop, and at zero the bot stops responding. Auto recharge removes that stop and buys more credits at $40 per 1,000, so with it on nothing caps the bill. Credits reset monthly and do not roll over, though recharged credits do not expire.
Additional watchouts
Credits plus add ons (auto recharge, branding removal, extra agents) push real cost well above the list price. Free and Hobby are undersized for production use. There is no visual flow builder, since agents are LLM driven only.
Overage / add-ons
Auto recharge tops up credits when the balance drops below a threshold, at $40 per 1,000 credits that do not expire, and manual add on packs run about $12 to $14 per 1,000. Without auto recharge, the bot stops responding at zero. Extra AI agents cost about $7 per agent a month, and choosing a premium model multiplies credit burn.
Sales call required
No, self serve available
Free / trial
Free plan with 50 message credits a month and 1 agent, deleted after 14 days idle
Lowest paid plan
Hobby at $40 per month, or $32 per month billed annually, with 700 message credits
Commercial notes
Agents train on websites, PDFs and Notion. AI Actions connect to Stripe, Calendly, Zendesk and Salesforce and handle order and cart operations, and a Shopify app is available. Chatbase works across OpenAI, Anthropic, Gemini, DeepSeek, Grok and other models, is GDPR compliant, does not train on customer data and hosts on AWS in the US.
Key ambiguities
Credit allotments and add on rates have changed. Real capacity depends entirely on model choice, and branding removal and custom domain pricing vary by tier.
Cancellation / refund
Self serve monthly or annual billing, with 20% off annual. Credits reset monthly and do not roll over, except auto recharge top ups, and Enterprise is on custom terms.
Support SLA / resale
Enterprise adds priority support, SLAs, a dedicated CSM, white labeling, a custom domain and higher limits.
Missing data
Add on rates for branding removal and a custom domain on each tier, and Enterprise pricing, which is custom.
Related vendors
- Decagon — AI agent platform purpose-built for customer support, with deep…
- 5.Y — 5
- Ada — Agentic customer experience platform whose AI agents autonomously…
- Aisera — Enterprise agentic AI platform that orchestrates specialized agents…
- ASAPP — Generative AI for enterprise contact centers that plugs into…
- Assembled — AI customer support orchestration platform that unifies autonomous…
Alternatives to Chatbase
The closest documented capability profiles to Chatbase among customer support agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Kommunicate10.0 / 14Matches Chatbase across all 14 documented capabilitiesChatbase vs Kommunicate →
- Botpress11.0 / 14Fuller documented coverage on Memory & State Persistence and Prebuilt Agents, Templates & PacksChatbase vs Botpress →
- LiveChatAI9.0 / 14A lighter documented profile than Chatbase
- BlueTweak9.5 / 14Adds documented Deployment & Data Residency
- Gallabox11.5 / 14Fuller documented coverage on Observability & Auditability and Prebuilt Agents, Templates & Packs
- NICE CXone10.5 / 14Adds documented Deployment & Data Residency
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded