GPT-4o

Claude 3.5

Gemma 4 · Ollama

Codex · Hermes

OpenCoder · Local

The intelligent chatbot platform · v0.9

Intelligent chatbots, built for the modern enterprise.

Connect any model. Upload your knowledge. Deploy in minutes — no glue code, no vendor lock-in. ModelPilot replaces your fragmented AI stack with one cohesive platform.

Get Started Free View Live Demo

500+ teams · 2.4M messages · 99.9% uptime

Connect any model Upload any knowledge Deploy in minutes Track every conversation $0 to start Open-source stack Connect any model Upload any knowledge Deploy in minutes Track every conversation $0 to start Open-source stack

Section 01 — The Manifesto

Most AI stacks are fragmented.
Ours is unified by design.
The product is the difference.

We believe building intelligent chatbots shouldn't require seven duct-taped tools. ModelPilot is the unified platform — knowledge, agents, analytics, widget — built for teams that move fast.

01 — Unified

One platform.
Any model.

OpenAI, Anthropic, Google, Groq, Mistral, Ollama — 8 providers behind a single LiteLLM gateway. Swap models without rewriting a line. Route by latency, cost, or availability.

02 — Contextual

Knowledge that actually retrieves.

Upload PDFs, scrape URLs, write FAQs. Auto-chunked at 256 tokens, embedded with OpenAI or Cohere, semantically retrieved through Qdrant. Real RAG, not keyword search.

03 — Measured

Every conversation, observable.

Token usage, cost per bot, resolution rate, sentiment score, conversation heatmap. Built into the platform — not a separate Datadog bill.

100% Free · Local · Unlimited

Codex & Gemma 4 —
unlimited local coding
via Ollama.

Plug in Codex, Gemma 4, Hermes, or OpenCoder — all running locally on your machine via Ollama. Zero API costs. Unlimited tokens. Full privacy. Paperclip your code, ship to OpenClaw, deploy with OpenCode.

Codex

Gemma 4

Paperclip Hermes

OpenClaw

OpenCoder

ollama — local

# Pull and run for free, forever
$ ollama pull gemma:4b
$ ollama pull codex
$ ollama pull hermes
$ ollama pull opencoder

# Wire to ModelPilot via LiteLLM
model_list:
  - model_name: codex-local
    litellm_params:
      model: "ollama/codex"
      api_base: "http://localhost:11434"

✓ Cost / month: $0.00
✓ Tokens / day: ∞

Section 02 — Use Cases

What teams build with it.

From customer support to internal knowledge bots — six patterns that ship on ModelPilot every week. Each one took less than an afternoon to deploy.

For Support Teams

Customer support automation

Answer FAQs, troubleshoot tickets, and escalate to humans when confidence drops. Replace expensive Intercom or Zendesk AI seats with a bot that knows your product.

78% ticket deflection $2k/mo saved

For Sales Teams

Sales qualification & booking

Qualify inbound leads, answer pricing questions, and book demos via Calendly. Sync captured leads to HubSpot or Salesforce automatically.

3.2× conversion lift 24/7 coverage

For Marketing & CS

Help center deflection

Drop in your docs, FAQs, or policy pages. The bot answers from your knowledge base with citations — no hallucinations, no off-topic replies.

335 chunks indexed 0.82 avg. similarity

⌘

For Internal Tools

Internal knowledge bots

Build employee-facing bots on your company wiki, HR policies, or engineering runbooks. Deploy to Slack or Teams with SSO.

SSO via Okta RBAC by team

文

For Global Products

Multilingual onboarding

Auto-detect user language, reply in 50+ languages with RTL support. Guide new users through product setup with interactive checklists.

50+ languages RTL supported

For Developers

Custom AI features via API

Use the REST API to power AI features inside your own product. Scoped API keys, webhooks on every event, SSE streaming for token-by-token replies.

OpenAPI spec SSE streaming

Section 03 — The Suite

Six modules.
One platform.

From knowledge ingestion to widget deployment to real-time analytics — every part of the chatbot stack, unified in a single product.

Featured

01 — Chatbot Builder

Any model.
Any persona.

Configure system prompt, temperature, fallback, and personality. Test live with real AI before deploying. Toggle human handoff, web search, and lead capture per bot.

GPT-4o

Claude 3.5 Sonnet

Gemini 1.5 Pro

Groq · Llama 3.1

6 chatbots · 5 personas

Real-time

02 — Analytics

Every conversation,
measured.

Message volume, cost per bot, sentiment heatmap, model distribution. Updates in real-time.

91%

Resolution rate · ↑ 2%

Live · 8,042 msgs today

RAG

03 — Knowledge Base

Semantic retrieval,
by default.

Drop PDFs, paste URLs, type FAQs. Auto-chunked, embedded, indexed in Qdrant. 335 chunks indexed across 6 documents — retrieved with 0.82 average similarity.

PDF

→

Chunks

→

Vectors

→

Context

Qdrant · OpenAI Embeddings

Orchestration

04 — AI Agents

Six agent types.
Zero code.

FAQ, Support, Sales, Onboarding, Language, and Handoff agents. Each with configurable confidence thresholds and escalation rules — powered by LiteLLM + Flowise.

FAQ Agent

Support Agent

Sales Agent

Language

Onboarding

Handoff

4 active · sentiment-aware

< 4KB

05 — Embed Widget

Vanilla JS.
Zero deps.

One script tag. Works on any site, CMS, or framework.

4KB gzipped · <50ms load

Webhooks

06 — Integrations

Webhooks &
6 connectors.

n8n, Slack, Zapier, Make, HubSpot, Zendesk — fire on any event.

conversation.ended

handoff.triggered

knowledge.indexed

7 event types

REST API

07 — API Access

Scoped keys.
Full control.

Generate API keys with granular scopes. Full curl reference, OpenAPI spec.

3 active · 48k req/mo

Section 04 — The Product

See it in motion.

A look at the actual interface — chatbot builder, analytics dashboard, and developer API. All shipping today.

Support Bot

v0.9 · GPT-4o · 342 conversations today

LIVE

Bot Name

Aria · Support Bot

AI Model

gpt-4o

System Prompt

You are Aria, a helpful support assistant. Be concise, accurate, and warm. Always check the knowledge base first.

Temperature · 0.65

Human handoff

Web search

Aria

● Online · GPT-4o

Hi 👋 I'm Aria, your support assistant. How can I help today?

How does the knowledge base work?

Upload PDFs, URLs, or FAQs — we chunk and embed them automatically. Your bot retrieves the most relevant context before every reply using semantic search through Qdrant.

What models do you support?

GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, Groq Llama, Mistral, Ollama — 8 providers total. Switch anytime in the model dropdown.

Type a message…

↑

Test Cost

$0.04 / conv.

Latency

820ms

Other Bots in Workspace

Aria · Support

GPT-4o · LIVE · 342 convs

99.9%

Sales Bot

Claude 3.5 · LIVE · 198 convs

99.7%

FAQ Bot

Gemini 1.5 · LIVE · 156 convs

99.4%

Onboard Bot

GPT-4o Mini · DRAFT · 0 convs

—

Recent Conversations● live

Support Bot

$0.04 Resolved

Sales Bot

$0.02 Active

FAQ Bot

$0.01 Resolved

Support Bot

$0.09 Handoff

Developer API

Open stack.
Full control.

FastAPI, LiteLLM, Qdrant — all open-source. Use the REST API, fire webhooks, or self-host the whole thing. Every part is inspectable, forkable, and fully self-hostable.

FastAPI

LiteLLM

Qdrant

Next.js 14

Supabase

Redis

Flowise

n8n

chat.py

rag.py

widget.html

Section 07 — Pricing

Start free.
Scale as you grow.

Flat platform fee. No per-token markup. You bring your own AI keys — we provide the infrastructure.

Starter

Forever free

✓3 Chatbots

✓5,000 messages / month

✓1 AI provider

✓Embed widget

✓Email support

Get Started Free

Common questions.

Do I need my own API keys?+

Yes — you connect your own API keys from OpenAI, Anthropic, Google, etc. You control costs directly with no token markup. ModelPilot charges a flat platform fee only.

Can I use open-source or free models?+

Absolutely. ModelPilot supports Ollama (local), Groq (free tier, 14k req/day), OpenRouter free models, HuggingFace serverless, and more. The entire stack can run at $0/mo.

How does the knowledge base work?+

Upload PDFs, paste URLs, or type FAQs. We chunk content into ~256-token segments, create embeddings using your chosen model (OpenAI or Cohere), and store them in Qdrant. At query time, the most relevant chunks are retrieved semantically.

Can I embed the widget on any website?+

Yes. The widget is vanilla JS, <4KB gzipped, zero framework dependencies. Paste one <script> tag before </body> — works on React, Vue, WordPress, Webflow, or plain HTML.

Is it self-hostable?+

The full stack is built on open-source tools (FastAPI, LiteLLM, Qdrant, Supabase, Redis) and is fully self-hostable. Enterprise plan includes a dedicated self-hosted deployment guide.

What about data privacy?+

Your data stays in your workspace. We're SOC 2 ready, GDPR compliant, and offer EU data residency on Pro/Enterprise. Conversation logs are encrypted at rest and never used to train models.

Intelligent chatbots, built for the modern enterprise.

Most AI stacks are fragmented.
Ours is unified by design.
The product is the difference.

One platform.
Any model.

Knowledge that actually retrieves.

Every conversation, observable.

Codex & Gemma 4 —
unlimited local coding
via Ollama.

What teams build with it.

Customer support automation

Sales qualification & booking

Help center deflection

Internal knowledge bots

Multilingual onboarding

Custom AI features via API

Six modules.
One platform.

Any model.
Any persona.

Every conversation,
measured.

Semantic retrieval,
by default.

Six agent types.
Zero code.

Vanilla JS.
Zero deps.

Webhooks &
6 connectors.

Scoped keys.
Full control.

See it in motion.

Open stack.
Full control.

Eight providers.
One API.

Start free.
Scale as you grow.

Common questions.

Your first bot,
five minutes away.

Intelligent chatbots, built for the modern enterprise.

Most AI stacks are fragmented. Ours is unified by design. The product is the difference.

One platform.Any model.

Knowledge that actually retrieves.

Every conversation, observable.

Codex & Gemma 4 —unlimited local codingvia Ollama.

What teams build with it.

Customer support automation

Sales qualification & booking

Help center deflection

Internal knowledge bots

Multilingual onboarding

Custom AI features via API

Six modules.One platform.

Any model.Any persona.

Every conversation,measured.

Semantic retrieval,by default.

Six agent types.Zero code.

Vanilla JS.Zero deps.

Webhooks &6 connectors.

Scoped keys.Full control.

See it in motion.

Open stack.Full control.

Eight providers.One API.

Start free.Scale as you grow.

Common questions.

Your first bot,five minutes away.

Most AI stacks are fragmented.
Ours is unified by design.
The product is the difference.

One platform.
Any model.

Codex & Gemma 4 —
unlimited local coding
via Ollama.

Six modules.
One platform.

Any model.
Any persona.

Every conversation,
measured.

Semantic retrieval,
by default.

Six agent types.
Zero code.

Vanilla JS.
Zero deps.

Webhooks &
6 connectors.

Scoped keys.
Full control.

Open stack.
Full control.

Eight providers.
One API.

Start free.
Scale as you grow.

Your first bot,
five minutes away.