The Ollabear platform — an AI agent that builds, reasons, and acts

Most "AI chat" at the small-business price point is a flow builder with a chatbot veneer. Ollabear is built the other way around: it's an agent that configures itself from your website, then reasons, calls your tools, remembers your visitors, and gets better over time. This page is the honest, complete picture — including what we don't do.

1. An AI builds your AI

Paste your website URL and a builder agent crawls your site, works out what your business does, and drafts a configured bot: a persona in your brand voice, a welcome message, and 6-10 business-specific FAQs from your real content. You confirm, embed one script tag, and you're live in about 90 seconds. No flow trees, no "upload your knowledge base" homework.

2. A real agent, not a script

The bot runs a tool-calling loop and speaks MCP (Model Context Protocol), so it can look things up and take actions — check an order, hit your own tools, use any of 280+ connected apps — not just recite text. That's the line between an agent and a chatbot, and it's the core of what we build.

3. Grounded in your content, safe by default

Answers come from your content via retrieval with citations, a confidence gate, and strictness modes — not a hallucinating ChatGPT wrapper. Every message passes a PII filter and a prompt-injection scorer before it's acted on.

4. It remembers, and it learns

Ollabear keeps per-visitor memory across conversations, so a returning visitor isn't met cold. And a learning loop clusters the bot's low-confidence answers, drafts better FAQ responses, and routes them to you for approval — the bot improves from its own gaps instead of staying static.

5. Knowledge-graph answers pilot

For structural, multi-hop questions — "which products ship to Canada and are on sale" — the bot can consult a per-bot knowledge graph built from your content, going beyond nearest-paragraph retrieval. This is live for pilot bots and rolling out; we don't claim every reply uses it yet.

6. 280+ integrations + open MCP

Connect Ollabear to 280+ apps no-code via our Activepieces piece, drive it from a scoped REST API, or let the agent use any MCP server as a tool. Automation in, actions out.

7. Geo-aware analytics analytics

Every conversation is enriched with the visitor's country, city and region, so you can see where demand is in your inbox and analytics. (Location-personalised answers are on the roadmap, not live — we flag the difference so the claim stays true.)

8. Multi-bot, isolated, and unlimited

One project equals one bot, with strict per-tenant data isolation and per-key domain + rate controls. Pricing is flat: unlimited bots and unlimited domains on every tier — only monthly conversation volume is metered. The free plan is genuinely free: 100 conversations a month, forever, no card.

What we deliberately don't claim

Honesty is part of the product. Ollabear is not end-to-end encrypted — the server needs plaintext to run RAG, guardrails, search and citations. It does not route between multiple LLM vendors inside the app (that's handled at the network edge). It is not SOC2 / enterprise-compliance tooling. If a vendor implies those things at this price, be skeptical.