From Documents to AI Agents in Minutes
Build powerful RAG agents without writing code. Connect your documents, configure your workflow, deploy in minutes.
Built for production stacks
Features
Everything you need to build AI agents
Ingest, orchestrate, and connect to your tools. Built for teams shipping real systems.
One-Click Ingestion
Drop any document - PDF, DOCX, images - and watch it become searchable instantly
No-Code Workflows
Build complex AI workflows with drag-and-drop. No coding required.
MCP Integration
First MCP-native RAG platform. Connect Claude and other AI tools seamlessly.
BYOK Model
Bring your own API keys. Control costs, keep your data private.
How It Works
1. Upload
Drop your documents - PDFs, DOCX, images, spreadsheets
2. Configure
Build your workflow with drag-and-drop. Add AI steps, filters, actions.
3. Deploy
Your AI agent is ready. Chat with it, embed it, or call via API.
Workflows
Pre-built templates to get started fast
Customize or build from scratch. Every workflow runs on LangGraph with full parallel execution support.
FAQ Bot
Auto-answer customer questions from your knowledge base
Support Agent
Triage tickets with context from docs and past conversations
Webhook Triage
Route incoming requests based on content classification
Content QA
Validate content against your brand guidelines automatically
Parallel execution built-in
Fan-out to multiple nodes simultaneously, then merge results. RAG search and web fetch run in parallel, reducing latency. Powered by LangGraph's Send API.
RAG Pipeline
RAG you can trust
Production-grade retrieval with the controls you need and the observability you deserve.
Quality
Hybrid + Rerank + Evaluation
- Vector + BM25 hybrid search
- Jina or LLM-based reranking
- Configurable score thresholds
- Built for evaluation workflows
Control
topK, threshold, configs
- Tune topK and score threshold
- Enable/disable query techniques
- Per-collection pipeline config
- A/B test retrieval strategies
Observability
Events, logs, health
- Real-time SSE execution events
- Per-node timing and outputs
- Execution history and replay
- Health checks and metrics
Pricing
Simple, transparent pricing
Start free. Scale as you grow. No hidden fees.
Free
For exploration and small projects
- 1 Collection
- 50 Documents
- 1 Workflow
- 100 Executions/month
Starter
For individuals and small teams
- 3 Collections
- 500 Documents
- 5 Workflows
- 1,000 Executions/month
- 1 MCP Server
- Webhook Triggers
Pro
For growing teams and production use
- 10 Collections
- 2,000 Documents
- 20 Workflows
- 5,000 Executions/month
- 5 MCP Servers
- All Triggers
- API Access
- 3 Team Members
All plans include a 14-day free trial. No credit card required to start.
Built with Trust & Transparency
We prioritize your security, control, and confidence in our platform
Packt Publishing AI Author
Built by an author with proven AI expertise
Your Data Stays Yours
BYOK model - bring your own keys for full control
MCP-Native RAG Platform
First RAG platform built for Model Context Protocol
Built on Open Standards
Leveraging LangGraph, OpenAI, and modern tooling
FAQ
Frequently asked questions
Everything you need to know about IgnitionAI.
IgnitionAI supports PDF, DOCX, PPTX, CSV, Excel (XLSX), JSON, JSONL, Parquet files, HuggingFace datasets, individual web pages, and full website crawls. We handle parsing, chunking, and embedding automatically.
Hybrid search combines vector embeddings (semantic similarity) with BM25 keyword matching. You can tune the alpha parameter to blend both approaches, or use Reciprocal Rank Fusion (RRF) for result merging. This gives you the best of both worlds: semantic understanding and exact keyword matching.
Yes! We currently support OpenAI (GPT-4, GPT-4o, GPT-4o-mini) and Anthropic (Claude 3.5 Sonnet, Claude 3 Opus). Enterprise plans can request custom model integrations.
Absolutely. All data is encrypted at rest and in transit. We use organization-scoped multi-tenancy, meaning your data is completely isolated from other customers. Authentication is handled by Clerk with JWT verification on every request.
Workflows can be triggered manually, via webhook (HTTP POST), or on a schedule (cron expressions with timezone support). Starter plans have manual + webhook, Pro and Enterprise get scheduled triggers.
Our workflow engine (built on LangGraph) supports fan-out/fan-in patterns. You can execute multiple branches in parallel (e.g., RAG search + web search simultaneously), then merge results before continuing. This significantly reduces latency for complex pipelines.
Not currently. We're a managed SaaS platform. However, Enterprise customers can discuss dedicated infrastructure options. Contact sales for more information.
Free plans get community support via GitHub. Starter plans include email support with 48h response time. Pro plans get priority email support with 24h response. Enterprise includes dedicated support with SLA guarantees.
Ready to build your AI agent?
Start free. No credit card required.
Join developers building the next generation of AI agents