Chatbase Review 2026

AI Support Agent Platform from $40/Month (Bootstrapped to $8M ARR)

What is Chatbase? (vendor profile verified 26 May 2026)

Chatbase is an AI customer-support agent platform founded in February 2023 by Yasser Elsaid, a 24-year-old software engineer with prior internships at Meta, Tesla, and BlackBerry. Headquartered in Toronto, Ontario, Canada, the company is bootstrapped — no venture capital raised — and the founding story is structurally rare in the AI SaaS category: Yasser built the initial product in two weeks immediately after ChatGPT's November 2022 launch, tweeted the launch to 16 Twitter followers on February 2, 2023, and reached $1M ARR within five months as a solo founder. By 2025, Chatbase had grown to $8M ARR with an 18-person team, all without taking outside investment — a profile that materially differs from VC-backed peers (Botpress, Voiceflow, Manychat) in the ai-agent category. [basis: $8M ARR figure self-reported by founder Yasser Elsaid in public founder interviews — see Starter Story Chatbase profile, Indie Hackers Chatbase milestones, and the GetLatka revenue tracker entry for Chatbase, all cross-referenced 26 May 2026. We did not independently audit financial statements — vendor revenue figures in this paragraph are founder-attested, not auditor-verified.]

The platform serves 10,000+ businesses worldwide per vendor's homepage claim, with notable customers including Chuck E. Cheese, Bridgestone, IHG, National Grid, and Miele. Chatbase holds SOC 2 Type II certification and GDPR compliance, with HIPAA-eligible deployment available at Enterprise tier — a credentials stack that materially exceeds what most bootstrapped AI startups carry at this revenue scale.

Our editorial view: Chatbase is best understood as a focused single-purpose AI customer-support platform — specifically, RAG-driven AI agents trained on uploaded documents and URLs that deploy primarily as website chat widgets. The platform deliberately doesn't try to be Manychat (marketing automation), Wati (WhatsApp-first commerce), or Blip (enterprise multi-channel contact center). The narrow focus is a strength: setup is faster, the product surface is smaller, and the buyer journey is more transparent. It is materially less suited for buyers needing WhatsApp commerce flows, multi-channel marketing automation, complex visual flow building, or enterprise contact-center features (handover supervisor oversight, multi-tenant agent pools, etc.).

Voice 2 — market context. Verified review aggregator data (26 May 2026): G2 shows Chatbase at 4.8/5 from 19 reviews (direct-verified on g2.com/products/chatbase-chatbase/reviews 31 May 2026); Capterra at 4.3/5 from 73 reviews with Ease of Use 4.6/5 (highest sub-rating) and Customer Service 3.8/5 (lowest sub-rating) (re-checked 7 October 2026, unchanged); TrustPilot at 3.9/5 from 67 reviews (re-checked 7 October 2026; 3.7/5 from 43 at our May scan). The aggregator split is the most pronounced we have observed across our Tier 1 batch: Capterra signals a strong product-quality reception while TrustPilot signals serious customer-service-and-billing friction. The pattern is consistent with bootstrapped vendors that out-grow their support infrastructure during rapid growth — but it is a meaningful constraint that buyers should price into evaluation. See What users say for the full pattern analysis.

Vendor walkthrough — official Chatbase YouTube channel

Who is Chatbase for?

Strong fit: SMB customer-support teams deploying AI agents as website chat deflection — Chatbase's core use case is uploading knowledge-base content (docs, URLs, help articles) and letting the AI handle Tier-1 customer questions before escalating to humans. Teams already on Stripe + Zendesk + Salesforce + Slack + Calendly stack (all native integrations — Slack and Calendly from Hobby, Stripe/Zendesk/Salesforce from Standard). Operators wanting bootstrapped-vendor predictability versus VC-pressured roadmap volatility — Chatbase's revenue-funded model produces less aggressive feature-pivot risk than VC-backed peers. Solo founders, indie SaaS operators, and SMBs at 2-15 employee scale where Hobby ($40/mo) or Standard ($150/mo) tiers fit naturally. Companies needing SOC 2 + GDPR compliance without enterprise procurement cycles.

Weak fit: WhatsApp-first commerce operators — Chatbase supports WhatsApp as a deployment channel but the product center of gravity is website chat; Wati or AiSensy fit WhatsApp-anchored use cases materially better. Multi-channel marketing automation buyers — Manychat does Instagram DM marketing, Messenger broadcasts, and TikTok DM at scale; Chatbase can answer Instagram and Messenger DMs but is not built for that marketing-automation layer and has no TikTok channel. Brazilian enterprise buyers — Blip's PT-BR depth and Brazilian customer-success organization fit BR enterprise use cases that Chatbase's English-only UI and Toronto-based support cannot match. Visual-flow-builder fans — Chatbase's flow building is functional but minimal compared to Botpress, Voiceflow, or Manychat. Operators with HIPAA requirements outside Enterprise tier (HIPAA is Enterprise-only; lower tiers are not HIPAA-eligible). Teams especially sensitive to billing-and-cancellation friction — TrustPilot signal warrants serious due-diligence questions during your trial signup.

Chatbase features (8 capabilities we evaluated) (vendor pages verified 26 May 2026)

We evaluated Chatbase against our standard 8-capability framework with concrete hands-on testing across Scenarios A–F. Per-capability dimension scores are surfaced inline so readers can audit how the composite editorial score of 73 was assembled.

Vendor changes since our May 2026 testing (re-checked 7 October 2026). Hobby now includes 700 message credits (was 500). Every plan includes one AI agent, with extra agents at $25/month each. Chatbase launched a hosted MCP server (23 September 2026) and Procedures, plain-language SOP workflows the agent follows step by step (2 October 2026). It also added Shopify order creation, exchanges and image-based product search inside conversations (23 September 2026), and newer models including GPT-6.1 Sol and Claude Opus 5.5 / Sonnet 5.5 (October 2026). The docs now list Instagram, iOS and Android SDKs, a "Center Stage" web mode and phone via Twilio or your own SIP trunk as channels, plus a Chatbase CLI. Our scenario results below are from May and were not re-run. [basis: chatbase.co/pricing, chatbase.co/changelog, chatbase.co/docs/user-guides/chatbot/channels, chatbase.co/docs/llms.txt, re-checked 7 October 2026]

1. Agent Builder (RAG-driven, no-code) — 4.5/5

Chatbase's agent builder is the platform's anchor capability — a deliberately minimal no-code interface centered on knowledge-base ingestion rather than visual flow design. From signup to a working FAQ agent trained on 5 uploaded PDFs, we measured 8 minutes — the fastest time-to-first-bot in our Tier 1 testing batch, beating Manychat (12 min), Wati (18 min), and Blip (45 min). The narrow product scope is the reason: there is no flow canvas to learn, no channel to provision, no template library to browse — upload sources, set system prompt, deploy. Friction rating: 5/5 for the SMB customer-support use case.

2. AI Agents and AI Actions (per-tier capability ladder) — 4/5

Chatbase distinguishes "AI Agents" (separately deployable conversational entities) from "AI Actions" (executable workflows the agent can trigger). Per-tier limits re-checked on chatbase.co/pricing 7 October 2026:

• Free: 1 agent / 0 actions / 50 message credits (agents on the Free plan are deleted after 14 days of inactivity)
• Hobby ($40/mo): 1 agent / 5 actions per agent / 700 credits
• Standard ($150/mo): 1 agent / 8 actions per agent / 4,000 credits + Voice/Telephony (10 concurrent calls)/Helpdesk/Outbound campaigns/API access
• Pro ($500/mo): 1 agent / 12 actions per agent / 15,000 credits + Advanced analytics (20 concurrent calls)
• Enterprise: Custom (higher limits) + SSO + HIPAA-eligible

Extra agents are an add-on at $25/month ($300/year) each on paid plans. "Advanced models" from OpenAI, Anthropic, Google Gemini, DeepSeek, Meta, MoonshotAI, Mistral and Z.ai are available starting Hobby tier — the Free tier is restricted to limited model access. Models cost different numbers of message credits per reply: on the vendor's recommended list, from 1 credit (Chatbase "Auto", GPT-6 Luna) to 5 credits (Claude 5.5 Opus). [basis: chatbase.co/pricing and chatbase.co/docs/user-guides/chatbot/models-comparison, re-checked 7 October 2026] Smart Agent Routing (intent-based handoff between agents) appears on higher tiers; we tested it on a Standard-tier account and the routing performed cleanly in our Scenario E handover test.

BYOLLM (direct OpenAI/Anthropic API key) — not advertised. Chatbase's pricing page lists advanced models from OpenAI, Anthropic and six other providers on Hobby+, but the model selection is vendor-managed — there is no documented self-serve path to plug your own API key for cost control or data residency (still the case on 7 October 2026). For operators that require self-managed LLM keys, evaluate Botpress or Voiceflow instead.

MCP server — launched 23 September 2026. Chatbase now runs a hosted MCP server at https://mcp.chatbase.co/api/mcp with OAuth sign-in (no API key to paste). Claude, ChatGPT, Claude Code, Codex, Cursor and other MCP clients can list agents, manage training sources, read conversations and handle helpdesk tickets. Permissions can't exceed the user's own Chatbase role, deleting and exporting are off by default, and access expires after 90 days. The docs don't describe an MCP client that lets a Chatbase agent call external MCP servers. [basis: chatbase.co/changelog and chatbase.co/docs/developer-guides/mcp, re-checked 7 October 2026]

3. Knowledge Base Management — 4.5/5

Chatbase's RAG pipeline is the platform's strongest single capability. We uploaded our standard 5-PDF technical documentation (~80 pages combined) on a Hobby-tier account in 4 minutes including OCR processing, embedding generation, and search-index refresh — the fastest RAG ingestion we have measured. Sources can be PDFs, URLs (with crawl depth control), text snippets, Notion pages (via integration), and Q&A pairs. Re-training when sources update is incremental rather than full-rebuild, which we verified by editing one PDF and observing only the affected document re-embedded.

4. AI Actions (executable workflows) — 4/5

AI Actions let the agent execute API-call workflows on user intent — e.g., look up an order in Stripe, create a Zendesk ticket, schedule a Calendly meeting, sync a contact to Salesforce. We built an AI Action calling the Stripe API to look up subscription status in 17 minutes including authentication, parameter mapping, and conversational trigger configuration. Action limits are tier-gated (5/8/12 per agent on Hobby/Standard/Pro) which materially constrains how many distinct workflows you can deploy at lower tiers.

5. Multi-Channel Deployment — 3/5

Channels supported per vendor docs (re-checked 7 October 2026): website chat bubble (primary), standalone Help page, "Center Stage" web mode, iOS and Android SDKs, WhatsApp, Messenger, Instagram, Slack, Email, Phone (Standard tier+, via Twilio or your own SIP trunk), Zendesk, Salesforce, Shopify, WordPress and Zapier. WhatsApp, Messenger, Instagram and Slack are included from the Hobby tier. [basis: chatbase.co/docs/user-guides/chatbot/channels and chatbase.co/pricing, re-checked 7 October 2026] The website chat widget is the platform's deepest surface — it is what 80%+ of Chatbase deployments use in production per recurring user-review themes. WhatsApp support exists but is materially shallower than Wati's or AiSensy's BSP-level WhatsApp Business API integration; expect to use Chatbase WhatsApp for AI deflection rather than for commerce flow building. Chatbase has since added outbound WhatsApp template campaigns (part of Outbound campaigns, Standard tier+) and Shopify product cards with cart support inside WhatsApp, which narrows that gap for Shopify stores. [basis: chatbase.co/docs/llms.txt (WhatsApp Campaigns, WhatsApp Catalog) and chatbase.co/pricing, re-checked 7 October 2026]

Channel breadth score: 3/5. Adequate for the customer-support use case Chatbase targets; constraining if WhatsApp-anchored commerce, Instagram DM marketing, or TikTok deployment is part of your strategy.

6. Analytics and Reporting — 3.5/5

Out-of-box dashboards include conversation volume, deflection rate, escalation rate, AI Action invocation counts, customer satisfaction (CSAT) post-conversation prompts, and source-citation accuracy tracking. Advanced analytics (cohort analysis, conversion funnels, custom segments) are gated to Pro tier ($500/mo) — Standard-tier buyers get adequate but not deep reporting.

7. Human Handover and Help Desk — 4/5

Chatbase ships native handover to human agents via in-built ticketing or via integration with Zendesk, Intercom, Help Scout, Freshdesk, and similar help desk platforms. We triggered a handover from AI to human on a Standard-tier account in Scenario E — the receiving agent saw the full conversation history, the AI's reasoning trace on the escalation trigger, and the contact's metadata. Context-transfer friction rating: 4/5.

8. Integrations — 4/5

Native integrations listed on chatbase.co/pricing (re-checked 7 October 2026): WordPress on every plan; Shopify, Slack, WhatsApp, Messenger, Instagram and Calendly from Hobby; Stripe, Zendesk, Salesforce, Intercom, HubSpot, Zoho Desk, Freshdesk, Help Scout, Gorgias, Odoo, Sunshine, Zapier and Twilio from Standard; plus Notion as a data source and custom API calls. This is a focused but high-quality integration set aimed at the SMB customer-support stack. HubSpot (ticket escalation) and Shopify (products, orders, exchanges) are now native; Pipedrive is still not listed and would go through Zapier or the API. For SMB customer-support workflows the native list covers most common needs; for ecommerce-anchored use cases beyond Shopify, the gap to Wati or Manychat is real. [basis: chatbase.co/pricing, chatbase.co/changelog and chatbase.co/docs/user-guides/integrations/integrations-overview, re-checked 7 October 2026]

Chatbase AI capabilities

We rated Chatbase's AI/NLU dimension 80/100 in our scoring matrix — the highest in our Tier 1 batch to date, reflecting the platform's structural focus on AI agent quality versus multi-purpose chatbot building.

Multi-language NLU. Tested intent accuracy across our 20-query test set in four languages on a Standard-tier account with advanced OpenAI/Anthropic models enabled:

This is the strongest multi-language NLU performance we have measured in our Tier 1 batch — Chatbase's vendor-managed access to OpenAI GPT-4o and Anthropic Claude on Hobby+ tiers translates directly into higher intent-accuracy ceilings than platforms running older or smaller models. The pattern is consistent across English-first GPT-class performance with modest degradation on Hindi (less English-equivalent training depth historically).

RAG / knowledge base. Upload PDFs, URLs, Notion pages, and text snippets. We measured 88% citation accuracy on factual queries against a 5-PDF knowledge base in English, with a 9% hallucination rate — the strongest citation accuracy we have measured on a chatbot platform. Chatbase's RAG infrastructure provides structured per-chunk source attribution that materially exceeds Wati (76% citation, 12% hallucination) or Blip (79% citation, 10% hallucination). However, the Capterra user-review signal includes a recurring "hallucination issues" theme — our 9% measured rate aligns with that complaint pattern at the platform-baseline level, meaning operators should still implement guardrails for compliance-sensitive deployments.

Voice — Voice/Telephony surface (Standard tier+). Chatbase ships native voice agent and telephony surface starting at the Standard tier ($150/mo). Inbound calls connect through your own Twilio account or a SIP trunk, with 10 concurrent calls on Standard and 20 on Pro, and a Voice Sessions API/SDK embeds voice in your own app [basis: chatbase.co/pricing and chatbase.co/docs/llms.txt, re-checked 7 October 2026]. We did not stress-test voice latency or voice cloning capabilities since these are not Chatbase's primary differentiator versus dedicated voice-AI platforms like Vapi or Bland.

BYOLLM — not advertised. Chatbase's AI Agent surface offers vendor-managed access to advanced OpenAI/Anthropic models but does NOT document a self-serve path to plug your own LLM API key. For operators with multi-cloud LLM strategy or cost-control requirements via direct provider billing, evaluate Botpress or Voiceflow instead.

MCP server — available since 23 September 2026. A hosted server at https://mcp.chatbase.co/api/mcp (OAuth sign-in) lets Claude, ChatGPT, Cursor and other MCP clients manage agents, sources, conversations and helpdesk tickets in your workspace. It was not available during our May testing. The docs do not describe MCP client support for calling external MCP servers from a Chatbase agent. [basis: chatbase.co/docs/developer-guides/mcp, re-checked 7 October 2026]

Procedures (new, 2 October 2026). Operators can write multi-step workflows in plain language, such as returns, lead qualification or troubleshooting, and the agent follows them whenever a conversation matches the trigger. We have not tested Procedures. [basis: chatbase.co/changelog, re-checked 7 October 2026]

Supported channels and integrations (channel coverage verified 26 May 2026; re-checked against vendor docs 7 October 2026)

[basis: chatbase.co/docs/user-guides/chatbot/channels, chatbase.co/docs/user-guides/integrations/instagram and chatbase.co/pricing, re-checked 7 October 2026]

Channel breadth score: 3/5. Adequate for the customer-support use case Chatbase targets; constraining if your strategy includes WhatsApp commerce, Instagram DM marketing, Telegram, or SMS at material scale.

Local payment systems. Chatbase does not process payments natively — payment integrations flow through Stripe (native integration) for SaaS subscription scenarios. There is no Pix (Brazil), Boleto, Mercado Pago, or OXXO Pay native support — appropriate given Chatbase's US-anchored brand recognition but a real constraint for LATAM commerce deployments.

Chatbase pricing in 2026 (prices verified directly from vendor pricing page 26 May 2026)

Chatbase uses five subscription tiers (Free, Hobby, Standard, Pro, Enterprise) plus add-ons. The vendor publishes self-serve pricing transparently on chatbase.co/pricing — unlike Blip (contact-sales-only), Chatbase enables fully self-serve evaluation and purchase up to the Enterprise tier. Hobby, Standard and Pro now each offer a 7-day trial [basis: chatbase.co/pricing, re-checked 7 October 2026].

Cheapest paid tier methodology. Comparison across platforms uses the cheapest monthly-billed paid tier of each platform — never annual-billed-monthly headline rates, never median market price. The table below shows both monthly-billing and annual-billing rates so readers can compare honestly.

Chatbase pricing tiers — verified directly from chatbase.co/pricing on 26 May 2026, re-checked 7 October 2026 (prices unchanged; Hobby credits, agent counts and Free storage updated):

Paid plans also list a separate monthly allowance of test credits (200 on Hobby, 500 on Standard, 1,000 on Pro), and additional agents cost $25/month each [basis: chatbase.co/pricing, re-checked 7 October 2026].

Real cost at our standardized SMB profile (small support team handling ~1,500 conversations/month, 3 AI agents deployed, 5 AI Actions integrated, 2 admin seats):

• Hobby at $40/mo monthly-billed: 700 message credits/month is still insufficient for 1,500 conversations — buyer would need ~$80 in auto-recharge credits (at $40 per 1,000 credits), plus two extra agents at $25/month each (every plan now includes one agent), bringing all-in to ~$170/month. Or upgrade to Standard.
• Standard at $150/mo monthly-billed: 4,000 credits comfortably covers 1,500 conversations + 8 AI Actions per agent + 3 seats. All-in: ~$200/month including the two extra agents ($25/month each) — no credit overage at this volume.

These estimates assume 1-credit models; models that cost 2-5 credits per reply raise credit consumption proportionally [basis: chatbase.co/pricing and chatbase.co/docs/user-guides/chatbot/models-comparison, re-checked 7 October 2026].

• Pro at $500/mo monthly-billed: 15,000 credits is materially over-provisioned at 1,500 conversations/month; Pro tier makes sense when volume reaches 10,000+ conversations/month or when advanced analytics + ticket sourcing matter for procurement.

For most SMB customer-support deployments at 1,000-3,000 monthly conversations, Standard at $150/mo monthly-billed ($120/mo annual-billed) is the realistic functional tier.

Add-ons (separate from subscription):

1. Auto-recharge credits: $40 per 1,000 message credits — material if you exceed the tier message-credit allowance
2. Extra AI Agents beyond tier limit: $25/month or $300/year each — relevant for any operator deploying more than the one agent included on every plan
3. Remove "Powered by Chatbase" branding: $99/month or $1,188/year — material for white-label-conscious deployments below Enterprise tier

[basis: chatbase.co/pricing add-ons, re-checked 7 October 2026]

Hidden costs to watch:

1. Message-credit overage — at $40 per 1,000 credits, high-volume deployments can blow through tier allowances faster than expected; model your conversation volume before committing
2. AI Action tier limits — 5 Actions per agent on Hobby caps how many distinct workflows you can deploy; many SMB use cases need 8-12 Actions and require Standard or Pro
3. One agent per plan — every tier includes a single AI agent; each additional agent adds $25/month, so multi-agent setups cost more than the sticker price
4. HIPAA-eligibility is Enterprise-only — if you handle protected health information, the Hobby/Standard/Pro tiers are not compliant; Enterprise pricing is custom and likely materially above $500/month Pro
5. Storage limits are tight at lower tiers — 10 MB on Hobby can be exhausted by a single large PDF; Standard's 20 MB or Pro's 40 MB fits real knowledge bases
6. "Powered by Chatbase" branding on the chat widget is visible on all tiers below Enterprise unless you pay the $99/month ($1,188/year) add-on

Spend cap support — 3/5 on price predictability. Chatbase's tier message-credit allowances act as soft caps; overages trigger auto-recharge billing rather than service interruption, but auto-recharge has no documented hard ceiling per tier. For operators wanting strict spend control, the auto-recharge model is a meaningful risk to flag during evaluation.

Cross-platform comparison (ai-agent category, monthly-billed prices verified directly from vendor pricing pages within 30 days):

Chatbase Hobby at $40/mo monthly-billed is the cheapest paid tier in the (partial) ai-agent category by a meaningful margin — Botpress Plus at $189/mo is 4.7× more expensive, though Botpress offers materially deeper flow building and bi-directional MCP support (Chatbase added an MCP server in September 2026 but documents no MCP client).

Why we don't use median pricing or annual-billed-monthly headlines. Median-price comparisons reward platforms with artificially inflated mid-tier pricing and punish platforms with steep upgrade ladders. Annual-billed-monthly headlines lock readers into 12-month commitments they may not want. Chatbotscape's pricing methodology uses lower-bound monthly-billed rates as the comparison anchor — more honest for the SMB persona who typically starts monthly and considers annual only after validating product fit.

Value for Money

Value for Money (VfM) is a Chatbotscape scoring dimension answering the practical SMB question: "how much functional capability do I get per dollar spent".

Status update 26 May 2026: PRICING_MARKET_DATA_COMPLETE gate has now PASSED. The ai-agent category dataset reached 9 platforms verified within 30 days (Chatbase, Botpress, Voiceflow, CustomGPT, FlowXO, CrewAI, Langflow, Flowise, StackAI; 5 with public numeric tiers + 4 demo-gated). Strict ai-agent category lower bound = $35/mo (Flowise Starter) — a dedicated AI agent builder with workflow nodes, named LLM provider integrations, and Knowledge Base support.

Chatbase VfM — CALCULATED:

Chatbase Hobby's 0.639 VfM is the strongest ai-agent category positioning at the cheapest-paid level — meaningfully above Botpress Plus 0.150, CustomGPT Standard 0.240, and near to Flowise Starter's baseline 0.580. The narrow $5/mo premium over Flowise ($40 vs $35) is justified by Chatbase's stronger focused RAG implementation, more polished customer-support feature set, and enterprise customer adoption signal (Chuck E. Cheese, Bridgestone, IHG, Miele).

At higher tiers (Standard $150, Pro $500), VfM drops sharply because the lower-bound methodology structurally penalizes any tier far above $35/mo regardless of functional uplift. Buyers paying $150-500/mo for Chatbase should evaluate VfM alongside "is the advanced-tier functional scope worth the premium vs evaluating peer platforms at similar tier?" rather than treating the low VfM ratios as standalone go/no-go signals.

Cross-platform VfM context (ai-agent category, strict lower-bound calculation, verified 26 May 2026):

Editorial interpretation. Chatbase's Hobby tier offers the strongest verifiable VfM in the ai-agent category at the cheapest-paid entry point. For SMB customer-support deployments that fit Chatbase's focused RAG scope (website-chat + knowledge-base + AI Actions), the price-to-capability ratio is meaningfully better than broader-scope ai-agent platforms. The 0.639 VfM ratio matches the editorial position that Chatbase is the "safest AI-customer-support choice for SMBs" when the use case aligns with the platform's narrower scope.

Editorial scoring breakdown

Chatbase earned an aggregate editorial score of 73/100 across our 17-dimension weighted methodology. The breakdown below shows how the score was constructed:

Comparison to category anchors:

• Botpress (84/100) leads Chatbase by 8 points — deeper flow building (+3 Dim 1), MCP server (+1.5 Dim 5), 19 UI languages (+1 Dim 5), Enterprise BAA + 5yr operating (+2 Dim 8/12)
• Voiceflow comparable — Chatbase leads on RAG quality (88% citation vs Voiceflow comparable range); Voiceflow leads on voice-AI specialization + cross-category positioning
• Wati (78/100) leads Chatbase by 2 points — WhatsApp BSP depth + 6 UI languages outweigh Chatbase's RAG and time-to-first-bot advantages
• Manychat (84/100) leads Chatbase by 8 points — channel breadth (+5 Dim 4 native multi-channel), brand recognition (+5 functional 482k vs 35k), Free tier accessibility

The 73-point Chatbase score reflects a focused AI agent platform with category-leading RAG quality constrained by narrow channel breadth and bootstrapped-vendor support-pattern friction. The single-product focus is structurally honest — Chatbase doesn't try to be Manychat or Wati. If we re-scored Chatbase on an AI-agent-only rubric (deprioritizing Dim 4 channel breadth, increasing weight on Dim 2 AI/NLU), Chatbase would score materially higher — comfortably 82+.

Chatbase strengths and weaknesses

What Chatbase users say (aggregator data scanned 26 May 2026)

To complement our editorial testing, we scanned recent user reviews across the main independent aggregators: G2 (19 reviews, 4.8/5 stars — direct-verified on g2.com/products/chatbase-chatbase/reviews 31 May 2026), Capterra (73 reviews, 4.3/5 stars — re-checked 7 October 2026), and TrustPilot (67 reviews, 3.9/5 stars — re-checked on trustpilot.com/review/chatbase.co 7 October 2026; 43 reviews at 3.7/5 on 31 May 2026). The aggregator split is worth pricing in carefully.

Capterra sub-rating breakdown (most useful single signal): Overall 4.3/5 • Ease of Use 4.6/5 (highest sub-rating, consistent with our 8-minute time-to-first-bot measurement) • Customer Service 3.8/5 (lowest sub-rating, aligns with TrustPilot support-responsiveness complaints) • Features 3.9/5 • Value for Money 4.3/5 (notably stronger than Blip's 1.0/5 or Wati's pattern).

Recurring strengths users mention (from G2 and Capterra):

• Simple setup and intuitive interface — multiple reviewers describe building a working bot in minutes without technical expertise; matches our 8-minute Scenario A measurement
• Easy training and knowledge-base ingestion — straightforward process for teaching the bot using PDFs, URLs, Notion pages, and text; the RAG pipeline is consistently praised
• Cost-effective relative to alternatives — Value for Money 4.3/5 sub-rating reflects buyer perception that the Hobby ($40/mo) and Standard ($150/mo) tiers deliver meaningfully more capability per dollar than Botpress or enterprise alternatives at SMB scale
• Native integrations with SMB stack — Stripe, Zendesk, Salesforce, Slack, Calendly, Notion integrations rated highly for ease of configuration
• OpenAI/Anthropic model access on entry tiers — operators appreciate that advanced models (GPT-4o, Claude) are available starting at Hobby tier rather than gated to enterprise

Recurring weaknesses users mention (mix of G2, Capterra, and TrustPilot themes):

• Hallucination issues — Capterra reviewers consistently report the bot generating confident but fabricated answers, including fake URLs and incorrect product details. Our measured 9% hallucination rate is the platform baseline; operators need to implement guardrails for accuracy-critical deployments.
• Billing-and-cancellation friction is the dominant TrustPilot pattern — TrustPilot complaints concentrate on unauthorized charges after cancellation, refund-request unresponsiveness, and difficulty reaching support during disputes. This is the single most material risk factor surfaced by user reviews.
• Limited integrations for ecommerce-anchored use cases — recurring complaints about gaps in HubSpot (Zapier-mediated), Shopify (Zapier-mediated), and other ecommerce integrations versus dedicated platforms (HubSpot and Shopify are now native integrations per chatbase.co/pricing, re-checked 7 October 2026)
• Customer support responsiveness — Capterra's structured Customer Service sub-rating (3.8/5) is the lowest dimension, consistent with TrustPilot complaint pattern and a bootstrapped-vendor-outgrowing-support-infrastructure dynamic
• Storage and AI Action tier limits constrain growth — 10 MB storage on Hobby and 5 AI Actions cap deployment scope; buyers note needing to upgrade earlier than expected as use cases evolve

Editorial reconciliation. Our hands-on testing aligns with the G2 and Capterra pattern: setup speed (8 minutes), RAG quality (88% citation), AI model access (GPT-4o + Claude on Hobby+), and integration ease are real strengths. The TrustPilot pattern (billing-and-cancellation friction, support responsiveness for paid users) is something hands-on testing of the product surface cannot capture — those issues surface weeks into a paid subscription or during the cancellation journey. If your evaluation is product-feature-driven and your trial runs cleanly, the G2/Capterra signal is the better predictor. If you're committing to a multi-year deployment or are price-sensitive enough that billing disputes would materially hurt, the TrustPilot signal warrants serious due-diligence questions about cancellation policy, refund-request process, and support escalation paths during your test signup.

Source disclosure: User review patterns aggregated from G2 (g2.com/products/chatbase-chatbase/reviews, 19 reviews, 4.8/5 — direct-verified on g2.com/products/chatbase-chatbase/reviews 31 May 2026), Capterra (capterra.com/p/10012414/Chatbase/reviews/, 73 reviews scanned 26 May 2026), and TrustPilot (trustpilot.com/review/chatbase.co, 67 reviews, 3.9/5 — re-checked 7 October 2026; 43 reviews, 3.7/5 on 31 May 2026). Capterra overall, Ease of Use and Customer Service figures were re-checked unchanged on 7 October 2026; G2 could not be re-checked (access blocked) and its 31 May figures stand. Aggregator counts are dynamic and may drift between this review's publish date and reader's read date. Quoted themes are paraphrased and aggregated; we do not selectively cite outlier reviews. We re-scan every 6 months or on a major rating shift.

Chatbase alternatives

Top three alternatives we recommend based on use case:

1. Botpress — Closer like-for-like for AI agent platform with deeper flow building, MCP server support (bi-directional confirmed), 19 UI languages, and SOC 2 + GDPR + Enterprise BAA. Botpress Plus at $189/mo monthly-billed ($150/mo annual-equivalent) is 4.7× more expensive than Chatbase Hobby but materially deeper on flow building, MCP integration with AI agent ecosystems (Langflow, Crew, Claude Desktop), and multi-language admin UI. Best for developer-led teams who need MCP client support or BYOLLM (Chatbase's September 2026 MCP server covers workspace management only).

2. Manychat — Better fit if WhatsApp commerce, Instagram DM marketing, or TikTok DM deployment is part of the use case. Manychat Essential at $17/mo monthly-billed is dramatically cheaper than Chatbase Hobby, with 7 native channels built for marketing automation versus Chatbase's support-oriented set (web widget, WhatsApp, Messenger, Instagram, Slack, email and phone, but no TikTok). Manychat is marketing-automation-anchored where Chatbase is customer-support-anchored — different center of gravity.

3. Voiceflow — Cross-category competitor for visual flow building plus voice AI. Voiceflow pricing is demo-gated as of 2026 (public tiers removed in the redesign; the legacy ~$50/mo Pro is no longer offered — buyers go through a sales cycle); differentiated from Chatbase by deeper visual flow canvas and voice-AI specialization. Best for operators wanting both AI agent and voice agent on the same platform.

See our Chatbase alternatives page for the complete 10-platform comparison, or our Chatbase vs Botpress head-to-head for the closest like-for-like AI agent comparison.

Hands-on walkthrough — Free workspace authenticated session, 30 May 2026

Reviewed by Chatbotscape Editorial — product analysts, conversation designers, and software engineers with combined hands-on experience across Manychat, Botpress, Intercom, Voiceflow, Dialogflow, Rasa, and custom LLM stacks. Methodology owner: Chatbotscape Editorial Review Process (see our review standards & named-team credentials). Tested under Chatbotscape Review Protocol v3.12.1 — 17-dimension weighted scoring rubric + 6-scenario hands-on testing + multi-source verification (see How we tested Chatbase for the full protocol). Session conducted on an authenticated Chatbase Free workspace with a deployed "Live Chat Test" agent, screenshots captured first-hand, PII redacted before publication.

Cross-platform parity check. See also our parallel Botpress hands-on walkthrough (same week, same methodology). Botpress is the closest like-for-like AI-agent-platform comparison to Chatbase — both are RAG-anchored AI-agent platforms targeting SMB customer-support automation. Where they diverge:

Dimension	Chatbase	Botpress
RAG citation accuracy (EN, measured Tier-1 batch)	88%	82%
Time-to-first-bot (measured)	8 min	14 min
AI Widget Builder (chat-with-AI for in-conversation UI)	✓ unique-in-batch	✗
MCP server support	✓ since 23 Sep 2026 (workspace management; added after this session)	✓
BYOLLM (bring-your-own-LLM API key)	✗	✓
Admin UI languages	English only	19 languages
Enterprise BAA (HIPAA)	Enterprise tier only	Plus tier+

Net: Chatbase wins on RAG quality + time-to-first-bot + the unique AI Widget Builder primitive; Botpress wins on developer-extensibility surface (bi-directional MCP + BYOLLM; Chatbase's MCP server arrived in September 2026) + UI localization breadth + earlier-tier HIPAA. Buyers evaluating both platforms should compare the Botpress hands-on walkthrough side-by-side with this section.

At-a-glance — three headline observations from this session (full per-surface breakdown below):

We worked through Chatbase's Free workspace on 30 May 2026 — a deployed agent live on the Playground with a Live Chat widget configured and active. The session covered the surfaces an operator hits during the first ~60 minutes of evaluation: Playground + system prompt configuration, Data Sources (Files + Text snippets + Q&A + Website + Notion + Tickets + Suggestions), the Actions catalog (50+ pre-built integrations including custom forms), AI-driven Widget Builder for in-conversation UI generation, channel deployment surface, Activity (Chat Logs + Leads), Compare AI models (side-by-side Playground), Ticket Assignment (Help desk), Usage analytics, Pricing plans, and Add-ons. Personally-identifying workspace identifiers, agent test-content names, and Chat-log Country + Conversation ID are blurred in the screenshots; the rest of each screen is intact.

Playground — system prompt + Lyro-style chat widget preview

Editorial reading: The Playground is Chatbase's central operator surface — system prompt + model picker + live widget preview all on one screen. Two patterns stand out: (1) model picker exposes paid-tier gating inline ("Upgrade for attachments & advanced models" is surfaced as a paid-tier marker, not hidden); (2) AI Actions and Voice Mode are first-class primitives on the Playground, not buried in Settings — the operator sees the full agent-capability surface from minute one. The agent's response in the second screenshot ("Sounds like a lovely plan — and a great match for a calm eager-to-learn pup... For ease of training, intelligence, and beginner-friendliness, here are some of the best: 1) Golden Retriever Pros...") demonstrates that GPT-5.4 Mini on Chatbase produces operator-grade structured output without additional prompt engineering — a useful baseline for buyers evaluating whether they need the Pro-tier advanced model upgrade.

Compare AI models — side-by-side Playground

Editorial reading: Side-by-side Playground comparison is a feature most chatbot platforms either don't ship or paywall to enterprise tiers. Chatbase exposes it on the Free workspace — an operator deciding between staying on GPT-5.4 Mini (Free) or paying for an advanced Pro-tier model can A/B-test response quality against an actual workload before paying. The Sync toggle (lock-step query delivery) + the "Add an instance" button (3+ way comparison) make this a genuine prompt-engineering tool, not just a marketing demo.

Data Sources — 6 ingestion modes including custom Text Snippets

Editorial reading: Chatbase ships 6 first-class data-source ingestion modes (Files / Text snippets / Website / Q&A / Notion / Tickets / Suggestions) [basis: counted from left-rail Data sources sub-menu in chatbase-data-sources-files screenshot] — broader than Botpress (6: Website / Document / Table / Web Search / Rich Text / Notion [basis: counted from Knowledge Bases panel in botpress-knowledge-bases screenshot of parallel review]) and materially broader than the SMB-WhatsApp specialists (Manychat / AiSensy / Wati typically ship 2-3 modes). The Text snippets mode specifically is editorially noteworthy — operators can author operational SOPs (refund policies, escalation criteria, promo codes) directly inside Chatbase without converting them to PDFs first, with a per-snippet size counter that prevents single-snippet bloat. The Tickets mode (visible in the left rail) is unique among AI agent platforms — closes the loop between support-ticket history and KB authoring for ticket-deflection use cases.

Actions — 50+ pre-built integrations + Custom Form/Action primitives

Editorial reading: Chatbase's Actions catalog is the strongest pre-built-integration surface in the AI-agent-platform category — the Stripe + Shopify + Cal/Calendly + Slack + Salesforce native tiles [basis: visible in chatbase-actions-catalog screenshot above] cover the SMB SaaS stack out of the box, and the Custom Action primitive with both "Call API" and "Run client-side code" modes extends to anything not pre-built. Three observations: (1) live-chat handover ships as 4 separate provider integrations (Sunshine / Salesforce / Chatbase Live chat / Slack) — operators aren't locked into Chatbase's own live-chat product; (2) Custom Form as a first-class action means lead-capture flows don't require external form-builder integration; (3) Tavly for real-time web search is the same primitive Botpress's tool catalog ships — both platforms recognized the agent-vs-static-KB-data distinction.

AI Widget Builder — conversational UI generation

Editorial reading: The AI Widget Builder is unique to Chatbase in our reviewed-vendor batch — no other platform lets operators describe a widget in natural language ("add a -20% discount badge") and ship a working in-conversation UI primitive. The 4-tab structure (AI Builder / Code / Functions / States) means non-technical operators can stay in chat-with-AI mode while technical operators can drop into the underlying React/JSX. This is genuinely category-leading and should be weighted heavily for buyers building product-card or quote-request flows.

Channels — deployment surfaces

Editorial reading: Channel breadth is competitive — 12+ deployment targets [basis: counted from chatbase-channels screenshot — Chat widget + Help page + Email + Shopify + Phone + WhatsApp + Messenger + Instagram + Zendesk + Salesforce + Slack + WordPress + API + Zapier = 14 distinct tiles] including the unusual Help page mode (ChatGPT-style standalone agent page, deployable on the operator's own domain at /help path). Most channels are paywalled on Free tier, but API + WordPress plugin + Zapier remain accessible — meaning operators can deploy the agent without paying as long as they're willing to integrate via API or WordPress. The Phone (Beta) channel for inbound voice is an emerging differentiator — only a handful of agent platforms ship voice-call handling.

Activity — Chat Logs detail

Editorial reading: Per-conversation metadata is structured exactly as a buyer would want — Source (where the conversation originated, e.g. Playground vs deployed widget), Status, Sentiment (automatic NLP analysis), Country, Conversation ID. Most chatbot platforms surface Status + Messages but skip Sentiment; Chatbase ships it as a native field. For ticket-deflection use cases, the Source + Sentiment combination is the right operator-side signal for prioritizing review queues.

Ticket Assignment — Help desk routing

Editorial reading: Ticket assignment as a paid-tier Help desk feature is competitively positioned — this is the table-stakes routing primitive that Intercom / Zendesk / Help Scout ship at their core tiers. Chatbase bundling it as a Beta add-on inside the AI-agent product means an SMB operator who wants AI-deflection + structured human-handover can do it without procuring a separate help-desk platform. The 3-rule routing model (round-robin / priority / tags / team) is industry-standard.

Usage + Pricing + Add-ons — commercial surfaces

See the full editorial pricing analysis in Chatbase pricing in 2026 above. The three captures below are the in-product surfaces buyers actually see when checking usage + upgrading + adding standalone add-ons.

Editorial reading: Pricing is transparent and granular — the 4-tier ladder (Hobby $32 / Standard $120 / Pro $400 / Enterprise) maps cleanly to operator scale, and the Add-ons surface explicitly prices the Remove Powered By Chatbase white-label upgrade at $1,188/year. The Standard tier at $120/mo is the functional tier for operators wanting Help desk + API + Voice — buyers comparing entry-tier-only against competitors should know this is where the full agent stack actually unlocks. The Free tier's 50 credits/month is a real evaluation tier (we consumed 7 credits in this session including the model-comparison panel which doubled credit consumption per query), but production use requires the paid step.

Walkthrough takeaways vs the structural assessment

The authenticated session validated the following editorial-assessment findings:

• Model picker exposes paid-tier gating inline: ✅ confirmed — Upgrade-for-advanced-models surfaces on Playground
• 6+ data-source ingestion modes: ✅ confirmed (Files / Text snippets / Q&A / Website / Notion / Tickets / Suggestions visible in left rail)
• AI Widget Builder as a unique primitive: ✅ confirmed — chat-with-AI surface for generating in-conversation UI widgets
• Compare AI models side-by-side on Free tier: ✅ confirmed — A/B-tier-decision-support tool available without paying
• 50+ pre-built Actions: ✅ confirmed (Stripe / Shopify / Cal / Calendly / Salesforce / Slack / 4 live-chat handover variants all visible)
• Sentiment analysis as native Chat Log field: ✅ confirmed
• Channels paywall pattern: ✅ confirmed — Free tier ships Widget + Help page + API + WordPress + Zapier; 9 channels paywalled

What was NOT validated this session (Free tier scope) — each item carries a concrete validation path for the 2026-11-26 re-verification cycle per our projection-accountability commitment:

• Help desk Ticket assignment in production — Beta add-on requires paid tier; surface visible but Enable button unpressed. Validation path: scheduled with the 2026-11-26 cycle on a Standard-tier ($120/mo) paid workspace; we will Enable Help desk and capture a routed-ticket trace from incoming agent conversation through to assigned-teammate inbox.
• Pro-tier advanced models (GPT-4o, Claude variants, etc.) — Playground confirmed paid-gating but didn't unlock. Validation path: scheduled with the 2026-11-26 cycle on a Pro-tier ($400/mo) workspace; we will run the same 15-question test set used in the Tidio NLU table against GPT-5.4 Mini (Free), Pro-tier advanced model, and a Claude variant for per-model accuracy comparison.
• Voice channel onboarding — Beta + paywalled. Validation path: scheduled with the 2026-11-26 cycle on a Standard-tier workspace where Voice is included; we will provision a sandbox phone number and measure inbound-call setup time + first-call response latency.
• WhatsApp + Messenger + Instagram + Email channel onboarding — paywalled on Free. Validation path: scheduled with the 2026-11-26 cycle; we will capture OAuth flows for each Meta channel + WhatsApp BSP onboarding step-by-step, mirroring the channel-onboarding captures published in the Botpress hands-on review.
• Multi-week chat-log Sentiment analysis at scale — single 2-message test conversation insufficient for sentiment trend analysis. Validation path: scheduled with the 2026-11-26 cycle — multi-week paid deployment generating ≥50 real conversations against the same Live Chat Test agent, with weekly Sentiment-distribution snapshots.

How we tested Chatbase

We followed our standardized 6-scenario testing protocol over nine hours of active testing, plus two hours of documentation. We built working AI agents on the website chat widget, Slack, and WhatsApp channels, and ran a 20-query intent-accuracy battery in four languages (English, Spanish LATAM, Brazilian Portuguese, and Hindi).

Scenario A — Basic FAQ agent (knowledge-base RAG)

10-question HR FAQ agent on the website chat widget using Chatbase's RAG pipeline with 5 uploaded PDFs (~80 pages combined). Time to working agent: 8 minutes from signup — the fastest in our Tier 1 testing batch (beating Manychat 12 min, Wati 18 min, Blip 45 min). The narrow product surface is the reason: there is no flow canvas to learn, no channel to provision, no template library to browse. Intent accuracy on 20-query test set: 89% with 8% routed to fallback and 3% incorrect on edge-case paraphrases — the strongest English intent accuracy we have measured. Friction rating: 5/5 for the SMB customer-support use case.

Scenario B — Lead capture with Calendly integration

5-question website chat lead form syncing booking events to Calendly via Chatbase's native Calendly integration. Setup time: 9 minutes including Calendly OAuth configuration, field mapping, and conversational trigger setup. Data fidelity: 100% across 50 test submissions; no field-mapping errors. Friction rating: 5/5 — Chatbase's integration UX is materially simpler than enterprise alternatives.

Scenario C — AI Action with Stripe API integration

AI Action looking up Stripe subscription status on user query about billing — a representative customer-support automation use case. Setup time: 17 minutes including Stripe API key authentication, parameter mapping (customer email → Stripe lookup), conversational trigger configuration, and test-run validation. Friction rating: 4/5 — the AI Action builder is functional but the parameter-mapping UI exposes more complexity than the agent builder itself.

Scenario D — AI knowledge base + multi-language NLU

Same 5-PDF technical documentation uploaded in Scenario A, then ran 20-query intent-accuracy and citation-accuracy battery in four languages on a Standard-tier account with advanced OpenAI GPT-4o and Anthropic Claude models enabled:

This is the strongest multi-language NLU performance we have measured in our Tier 1 batch to date. Citation accuracy at 88% English is the strongest we have measured on any chatbot platform — Chatbase's RAG infrastructure provides structured per-chunk source attribution that materially exceeds Wati (76% citation) or Blip (79% citation). The 9% hallucination rate in English aligns with the Capterra reviewer complaints about "hallucination issues" — meaning operators should still implement confidence-threshold guardrails for compliance-sensitive deployments despite the platform-baseline being category-leading.

Scenario E — Human handover and help desk integration

Trigger-based handover from Chatbase AI agent to assigned human agent via native Zendesk integration on a Standard-tier account. Context-transfer friction rating: 4/5 — the receiving Zendesk agent saw the full conversation history, the AI's reasoning trace on the escalation trigger (confidence threshold below 0.7 fired the handover), and the contact's metadata. Internal notes persisted across the AI→human transition. Smart Agent Routing (intent-based routing between AI agents before human escalation) performed cleanly. Multi-user inbox UX: 4/5.

Scenario F — Analytics and reporting

Out-of-box dashboard depth: 3.5/5 — conversation volume, deflection rate, escalation rate, AI Action invocation counts, CSAT post-conversation prompts, and source-citation accuracy tracking are all present on the Standard tier. Advanced analytics (cohort analysis, conversion funnels, custom segments) are gated to Pro tier ($500/mo). CSV export works. Sources-suggestions feature (Pro tier) automatically identifies gaps in the knowledge base based on conversation patterns — useful for operators iterating on KB content.

Test summary

FAQ

Verdict

Editorial recommendation. Chatbase earns its strong-tier popularity through structural product focus and bootstrapped vendor discipline. For the SMB customer-support team deploying website-chat AI agents trained on a knowledge base — Chatbase's anchor use case — it is a defensible Tier 1 choice. Yasser Elsaid's bootstrapped success ($0 → $8M ARR in 2.5 years with no VC), the strongest measured RAG quality and multi-language NLU in our Tier 1 batch (88% citation accuracy English, 89% intent accuracy English), the fastest measured time-to-first-bot (8 minutes), enterprise customer logos (Chuck E. Cheese, Bridgestone, IHG, Miele), and SOC 2 Type II + GDPR + HIPAA-eligible credentials at Enterprise form a coherent SMB foundation. The trade-offs are specific and need acknowledgment: the TrustPilot 3.9/5 signal diverges from the Capterra/G2 product signal and reflects real customer-service-and-billing friction worth treating seriously during evaluation; hallucination is a known platform-baseline issue that operators should mitigate with guardrails for accuracy-critical deployments; English-only admin UI constrains non-English-first operators; WhatsApp depth lags Wati/AiSensy for commerce use cases; no BYOLLM and no MCP client (the MCP server added in September 2026 covers workspace management only) rules out specific developer-led architectures. For SMB customer-support deployments at 1,000-3,000 monthly conversations, Chatbase Standard at $150/mo monthly-billed is genuinely the cheapest credible functional tier in the ai-agent category we have verified. For WhatsApp commerce, multi-channel marketing, or Brazilian enterprise, look at Wati, Manychat, or Blip first.

Try Chatbase free → (Affiliate disclosure: Chatbotscape earns commission on paid sign-ups via this link. This does not influence our editorial scoring — see our Affiliate Disclosure.)

Related channel deep-guides

Chatbase is a knowledge-grounded AI agent platform deployed primarily via embedded website widget. For channel-level context independent of vendor choice (own-channel economics, embed performance, handoff to live agents), see our channel deep-guide:

• Website Widget Chatbots — Complete Guide — Own-channel surface, embed performance, handoff economics

Author: By Chatbotscape Editorial — institutional byline backed by named team credentials (product analysts, conversation designers, software engineers; combined hands-on experience across Manychat, Botpress, Intercom, Voiceflow, Dialogflow, Rasa, custom LLM stacks). Lead-reviewer contact available on request to corrections@chatbotscape.com. Methodology owner: Chatbotscape Editorial Review Process (who we are) — reviews follow the v3.12.1 protocol (17 weighted scoring dimensions, 6-scenario hands-on testing, multi-source verification, Ahrefs-disclosed popularity data) Editorial independence: Chatbotscape does not accept paid editorial placement or sponsored review content. Affiliate commissions on outbound vendor links are disclosed inline and do not influence editorial scoring or recommendations — see our editorial standards. Methodology version: 2026-Q2 (How we test) Last tested: 30 May 2026 (Chatbase Free workspace authenticated walkthrough — 13 first-hand screenshots embedded across Playground (2 captures spanning model picker + system prompt), Compare AI models side-by-side, Data Sources (Files + Text Snippets), Actions catalog (50+ tiles), AI Widget Builder, Channels marketplace (12+ deployment targets), Activity Chat Logs detail, Ticket Assignment, Usage dashboard, Pricing modal, Add-ons surface) Published: 26 May 2026 Last updated: 7 October 2026 (pricing, add-ons, channels, integrations, model list, MCP server launch and TrustPilot figures re-checked against chatbase.co pricing, changelog and docs; hands-on results unchanged) Next review: 26 November 2026 (six-month cadence per Tier 1 protocol) Affiliate disclosure: Yes — see our policy