Bhotforge
Log inGet started
v4.2Now with one-click WhatsApp + Slack deploys →

The knowledge base that talks back in a sprint.

Bhotforge turns your docs, PDFs, and help center into an AI support agent with zero hallucinations. We lift the heavy n8n plumbing, vector DBs, and model routing — so anyone on your team can ship one. No ops team required.

Start free, no card Watch a 90-sec demo
10-minute setupClaude · GPT-4o · GeminiSOC 2 Type II
acme.com / support-agentlive
Visitor · 2s ago
How do I migrate my workspace to the team plan mid-cycle?
Bhotforge · trained on 412 docs
You can upgrade from your Billing settings — we prorate the remaining month, and team seats activate instantly. Existing API keys keep working; no code changes needed.
grounded in:billing/upgrades.mdapi/auth.mdx
Visitor · now
Perfect, what about the invoice?
Bhotforge
Ask about billing, docs, or features...
retrieval latency
184 ms
resolution rate
92.4%
▲ 8.2%this week
Built for teams that ship
Response timeHours → seconds
Setup~10 minutes
HallucinationsStrictly grounded
Uptime target99.99%
Capabilities

Everything a RAG stack needs.
Nothing you have to glue together.

Skip the weekend of wiring vector DBs to workflow tools. Bhotforge ships the whole pipeline — retrieval, routing, observability, auth — behind a single dashboard.

Knowledge Autopilot. Your data is your agent's brain.

Upload PDFs, sync URLs, or connect a Notion workspace. Hybrid search + semantic reranking keeps answers honest — every response cites the chunk it came from. Auto re-indexes as your docs evolve.

query
→
embed
→
retrieve
→
rerank
→
generate
pgvector1,284 chunks indexed● ok
bm25hybrid weighting 0.6 / 0.4● ok
cohere-rerank-v3top-k = 8 → 3● ok

Multi-model. One switch.

Claude 3.5 Sonnet, GPT-4o, Gemini 1.5 Pro — or bring your own keys.

Claude 3.5
Sonnet
↔
GPT-4o
+ Gemini

Be where your customers are.

Web widget, React hook, REST, WhatsApp Business, or Slack — one agent, every channel.

sfofra

Sub-second p95.

180ms

Document to API in 10 minutes.

REST API, TypeScript SDKs, row-level security. SOC 2 Type II. Your data never trains anyone else's model.

  • ● SOC 2 Type II
  • ● Row-level security
  • ● BYO encryption keys

See every turn of every conversation.

Live traces show retrieval hits, model calls, tokens, and latency — grouped by conversation.

How it works

From zero to live agent. Under 10 minutes.

No coding, no infrastructure, no AI expertise. If you can upload a file and click a button, you can launch a support agent that actually works.

01

Upload your content

PDFs, web pages, Notion — just drag and drop. We index everything automatically so your bot knows your business inside-out.

Works with any file formatReady in under 60 seconds
02

Set the personality

Pick a tone, choose a model, set boundaries. Use a ready-made template or write a simple description — no technical jargon needed.

Templates for support, sales & HRZero config required
03

Go live instantly

One line on your website, or flip a switch for WhatsApp and Slack. Your customers get instant answers — you get your time back.

Website, WhatsApp & SlackCopy, paste, done
04

Improve on autopilot

Your bot gets smarter every day. Review conversations, approve corrections, and watch resolution rates climb — automatically.

Self-improving knowledge base92% avg. resolution rate
Pricing

Free forever for small teams. Fair as you scale.

Free
$0/mo
For solo builders testing the waters. No credit card, ever.
  • ✓1 chatbot
  • ✓10 documents
  • ✓500 messages / mo
  • ✕WhatsApp channel
  • ✕Bring your own API keys
Get started
Recommended
Pro
$24/mo
For growing teams that need channels, guardrails, and analytics.
  • ✓5 chatbots
  • ✓100 documents
  • ✓10,000 messages / mo
  • ✓WhatsApp + Slack
  • ✓Bring your own API keys
Start 14-day trial
Enterprise
$79/mo
For scale, compliance reviews, and a dedicated solutions partner.
  • ✓Unlimited bots & docs
  • ✓SSO / SAML + audit logs
  • ✓Dedicated account team
  • ✓Custom data residency
  • ✓99.99% uptime SLA
Contact sales
FAQ

Questions, answered.

Do I need to know how RAG works to use this?
No. If you can drop files into a folder and write a sentence describing your bot's job, you're qualified. We handle chunking, embedding, retrieval, reranking, and model routing under the hood — none of which you need to configure to get a production-ready bot.
How is this different from building it in n8n or LangChain myself?
You skip the plumbing. A typical n8n RAG build is 12–14 nodes: ingestion, chunking, vector store, retrieval, reranker, LLM, memory, logging, auth. Bhotforge ships that pipeline tuned and observable out of the box. You still get escape hatches — webhooks, custom tools, BYO models — but the default path is ten minutes, not ten days.
Which LLMs can I run it on?
GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, and any OpenAI-compatible endpoint. You can use our shared pool or bring your own keys — per bot, per environment.
Is my data used to train anyone else's model?
Never. Documents live in a dedicated vector namespace with row-level security. We're SOC 2 Type II and support BYO encryption keys on Enterprise.
What happens if the bot doesn't know the answer?
By default it says so, and offers to route the user to a human channel you define. You can configure fallback prompts, Slack escalation, or a ticket webhook — per bot.
Ready when you are

Your first bot is ten minutes away.

Free forever for small teams. Upgrade only when it's paying for itself.

Start building — free Book a 15-min demo
No credit card10-min setupCancel anytime