The open-deployable
AI support platform
Upload your knowledge base, deploy to any channel, keep your data on your terms. RAG-powered chatbots with enterprise-grade human handoff — running on your hardware or ours.
Enterprise power, without vendor lock-in
Everything a modern support team needs — from AI to human handoff — in one deployable platform.
Your data, your infrastructure
Deploy on-premise with Ollama + pgvector for zero external API calls. Fully air-gapped option available — your knowledge base never leaves your servers.
One knowledge base, every channel
Web widget, WhatsApp, and Telegram — all managed from a single admin panel. One upload, consistent answers everywhere.
Human-in-the-loop, built in
Confidence-scored escalation automatically routes complex issues to live operators. Appointment booking included — no third-party integrations needed.
Everything in one platform
No stitching together tools. Kiedo ships with the entire stack — from RAG pipeline to analytics dashboard.
Streaming WebSockets
Real-time chat for visitors and operators via dedicated WebSocket endpoints. No polling, no latency.
Pluggable LLM Providers
Claude Haiku 4.5 by default. Switch to Ollama phi4-mini, llama3.2, qwen3, or ministral — no code changes required.
Analytics Dashboard
Conversation telemetry, sentiment analysis per turn, and Grafana dashboards provisioned on deploy.
2FA & JWT RS256
TOTP two-factor authentication via QR code. Argon2 password hashing. Encrypted tenant secrets at rest.
Multi-Tenant White-Label
Agencies and SaaS platforms can resell Kiedo under their own brand. Per-tenant personality customization and global prompt library.
OpenTelemetry Tracing
OTel spans on embedding, bot engine, and DB calls. Per-turn bot_traces telemetry table out of the box.
Document Ingestion
PDF (PyMuPDF), DOCX, TXT, MD, HTML, CSV — up to 50 MB per file. Live progress polling during reprocessing.
Flexible Billing
Stripe, crypto, manual, or simulated billing providers. Subscription plans, promo codes, superadmin management.
i18n-Ready Widget
Internationalization-ready widget bundle. Deploy to global audiences from day one with locale support built in.
Designed for reliability at scale
Seven Docker-composed services, all orchestrated and observable from day one.
Meet your customers where they are
One knowledge base. Three channels. Zero duplication.
Web Widget
Embed with a single async script tag. Configurable personality, i18n-ready, appointment booking built in. Deploy in under a minute.
See embed code →Connected via Meta Cloud API with webhook support. Your existing WhatsApp Business number, supercharged with AI and escalation.
Setup guide →Telegram
Built on aiogram for reliable async message handling. Full support for escalation, sentiment analysis, and all Kiedo features.
Setup guide →Simple, transparent pricing
Start free, scale with confidence. Every plan includes the full RAG pipeline and multi-channel delivery.
For small teams and startups exploring AI support.
- 1 tenant / 1 bot
- Web widget channel
- PDF, DOCX, TXT ingestion
- 1,000 AI conversations/mo
- Claude Haiku 4.5 LLM
- Basic analytics
For growing teams that need multi-channel and operator tools.
- 10 tenants / unlimited bots
- Web, WhatsApp & Telegram
- All document formats + 50 MB limit
- 10,000 AI conversations/mo
- Human escalation + bookings
- Full analytics + OTel traces
- Grafana dashboards
- 2FA + encrypted secrets
Full control, full compliance, zero third-party exposure.
- Unlimited tenants & bots
- Self-hosted Docker Compose
- Ollama on-prem LLM (air-gapped)
- Unlimited conversations
- White-label reseller rights
- SLA + dedicated support
- Custom SSO / SAML
Common questions
Yes. Set ALLOW_OLLAMA_IN_PRODUCTION=true in your environment and Kiedo will use a locally-running Ollama instance for both embeddings (mxbai-embed-large) and inference (phi4-mini, llama3.2, qwen3, ministral-3b). Zero external API calls — fully air-gapped deployments are supported.
The RAG engine scores each response with a confidence value. When confidence drops below the configured threshold, the bot emits an <ESCALATE reason=...> tag that triggers a real-time operator notification via the /ws/admin WebSocket. Operators can then take over the conversation instantly from the admin panel.
Kiedo ingests PDF (via PyMuPDF), DOCX, TXT, Markdown, HTML, and CSV files. Each file can be up to 50 MB. Uploaded documents are chunked with a recursive character splitter (300 tokens, 30-token overlap, configurable) and indexed into PostgreSQL with pgvector.
Most chatbot platforms are single-tenant, web-only, and lock you into their hosted infrastructure. Kiedo is multi-tenant, supports three channels from one knowledge base, and can be self-hosted with zero external API dependencies. You can also swap the LLM, embedding model, and payment provider independently — no replatforming required.
The SaaS version is ready immediately. For self-hosted, a standard docker compose up deploys all 7 services (FastAPI, PostgreSQL, Redis, admin panel, nginx, Grafana, and the widget bundle) in minutes. Full documentation is available for production hardening and SSL configuration.