Skip to Content
Introduction

NavFlow

NavFlow is the data layer for your AI agents. It receives events from any source — OTLP telemetry, HTTP/JSON webhooks, or any system that can POST JSON — then triggers, transforms, and routes them to your AI agents for classification and enrichment. Results are delivered to downstream sinks like Slack and webhooks.

Sign up for free at app.navflow.ai  and start processing up to 10,000 events/month at no cost.

What it does

  • Receives events from OTLP (logs, traces, metrics), HTTP/JSON webhooks, or any JSON source via a high-throughput data plane
  • Triggers and transforms events using expression-based rules before they reach the agent
  • Batches and forwards matching events to your AI agents for processing
  • Context windows — accumulate events in a sliding window and trigger the agent with full temporal context, so your agent sees the trail of events leading up to a trigger — not just isolated alerts
  • Routes enriched output from agents to configured sinks (Slack, webhooks)
  • Manages everything through the NavFlow dashboard and API

Key features

  • Context windows with sliding triggers — the standout capability. Accumulate events in a sliding time window, trigger agent invocation based on expression rules, and deliver the trigger event plus its full context window to the agent. Configure trigger expressions, context filters, window durations, and group keys to give your agent temporal awareness without writing any state management code.
  • Separate control and data planes — management API stays isolated from the high-throughput data path
  • Pluggable AI agents — agents are standalone services (Python, Go, any language) that receive events and send results back via the NavFlow SDK
  • Per-project pipelines — each project has its own filter rules, transforms, agent endpoint, and sinks
  • Durable streaming — NATS JetStream backbone with exactly-once message delivery between pipeline stages
  • Expression-based filtering — drop irrelevant events before they reach your agent, reducing noise and LLM cost

Getting started

  1. Getting Started — sign up and send your first events
  2. Architecture — understand the system design and data flow
  3. Configuration — set up pipelines, filters, transforms, context windows, and sinks
  4. Agents — build and connect your own AI agent
Last updated on
Getting Started